Selling vs Stacking Your Staking Emissions: What the Data and Real Investor Behavior Show
By Rick aka 'TaoIsTheKey' on X, founder of taoemissions.com. Published 21 November 2026, revised the same day to measure everything in US dollars. Data runs to 20 November 2026.
The short version
If you stake in Bittensor subnets, you get paid every day, and you have one choice to make about that pay. You can stack it, which means leaving it staked so it compounds. Or you can sell it every day: sell the day's emissions into TAO, then sell that TAO for USDC (a dollar stablecoin) on an exchange, where it earns interest. Only the second step locks the profit in dollars. TAO kept as TAO still falls when TAO falls.
We tested both on real Bittensor data, in dollars. Over the last six months, selling daily to USDC beat stacking in 80 of 90 random 10-subnet portfolios, and over three months it beat stacking in all 90. It had the smallest worst drop in every one of those portfolios.
Stacking does make more money when prices rise and you sell near the top. Our three model portfolios rose against TAO over six months, and stacking won there. The research on real crypto investors shows that far more people sit on big gains than sell them, and around half hold all the way through a crash. When we judged both strategies at the moments real people actually sell, selling daily to USDC came out ahead in five of the six full-cycle cases we modelled, and it cut the gap between selling at a good time and selling at a bad time roughly in half.
The conclusion is clear: for the way most people actually invest, selling your emissions every day and taking them to dollars beats stacking. Stacking only comes out ahead for the investor who sells at or near the top, and even then the trade is poor: in every case we tested, stacking added more risk than it added reward, between $1.18 and $1.81 of extra swing for every extra dollar of upside in the full cycles. Selling daily limits the upside somewhat and limits the downside far more. Section 10 lays out the full case.
1. Why I care about this
In March 2024 my crypto portfolio was worth about $2.6 million. I didn't sell. I believed the four-year cycle theory and expected the bull market to run deep into 2025, with 2026 as the bear market. Most of my money was in TAO, and TAO peaked on 7 March 2024 at about $757 (CoinGecko). It never got back to that high. Bitcoin kept going and topped in October 2025, but most altcoins never got the late run they had in past cycles: almost two-thirds of 2021's top 100 altcoins are down more than 90%, while Bitcoin is up about 30% over the same stretch (CoinMarketCap research, reported by Fortune). So March 2024 turned out to be the high point of my portfolio in that bull run, and it was pretty much down from there. Today it's worth about $400,000, roughly 85% below that high. I barely took any profits on the way up.
I knew how to buy and had no plan to sell, so I kept holding on conviction. I built taoemissions.com so stakers can sell their emissions from their own wallet on a fixed schedule and take the money off the table in dollars, without anyone having to call the top.
2. A few terms
Emissions. When you stake TAO into a Bittensor subnet, you receive that subnet's own token (called its alpha). Your stake earns new alpha every day, and those daily rewards are your emissions.
Validator. The operator you stake through. It keeps a cut of the rewards and passes the rest to you. Taoemissions.com names a best validator on each subnet: the highest-paying one with no warning signs, such as a very small or brand-new stake. Of the 92 subnets where we named one, 79 paid 30% a year or more on 20 November 2026, and the middle subnet paid about 40% a year. Those rates are paid in the subnet's own token, and that token's price moves.
USDC. A stablecoin that tracks the US dollar one for one. Exchanges such as Coinbase pay interest on USDC balances. I earn 3.75% a year on mine at Coinbase, so that's the rate this study uses. Coinbase sets the rate by account and can change it each month.
Worst drop. How far a holding fell from its highest point at any time during a period, measured in dollars. Traders call it the maximum drawdown.
Realized. A gain or loss is realized when you lock it in by selling.
The two ways to handle emissions:
- Stacking leaves the rewards staked, so your token count compounds. This is what happens if you do nothing.
- Selling daily to USDC sells each day's rewards into TAO, sends that TAO to an exchange, and sells it for USDC the same day. Your original stake stays the same size, and a pile of dollars builds up beside it, earning interest. Taoemissions.com handles the first step from your own wallet, and you do the second on your exchange.
Two more cases show up for comparison. No staking means buying the tokens and just holding them, which is how people hold most of crypto, because most tokens pay nothing to hold. Selling to TAO and keeping the TAO is the halfway version: it gets you out of the subnet token but leaves you fully exposed to TAO's own price.
3. What the research says about how crypto investors really behave
We pulled every credible source we could find across all of crypto. Here's what it supports, and where it runs out.
Far more people sit on gains than sell them
No study follows a group of crypto investors from a market top to the next bottom and reports where each person sold. The closest evidence:
- The newest US government survey is a staff working paper from the SEC's Office of the Investor Advocate. It doesn't speak for the Commission. It was published in July 2026 and covers 5,550 US adults surveyed in July 2025, when Bitcoin was close to its high. Among current crypto owners, 37.1% said their crypto "gained a lot, but [they] didn't sell," 30.6% said it "lost a lot, but [they] didn't sell," and only 11.0% said they "sold and made a lot of money." More than three times as many owners held through a big gain as sold and made a lot.
- A study of about 200,000 eToro accounts found that the same people who trim winners and buy dips in stocks and gold don't do it in crypto. They hold through big price swings and don't trim their crypto when it runs up sharply (Kogan, Makarov, Niessner and Schoar, NBER Working Paper 31317, published in the Journal of Financial Economics in 2024).
- On the Bitcoin chain, at least 15% of all coins changed hands within about 14% of the November 2021 top and were still sitting with their buyers three weeks later (Glassnode, Week 47 2021). That figure counts coins and says nothing about how many people sold, and the true amount traded near the top was higher. It does show how much was bought close to the peak.
Around half hold all the way through a crash
- Two surveys taken in mid-2022, during the crash and before its final low, found that 54% (Civic Science) and 55% (Appinio) of crypto investors held through it without selling. Neither gives a sample size.
- In the UK regulator's survey series, 45% of crypto users reported a net loss in 2022 and 32% reported a gain (Financial Conduct Authority and YouGov).
- At the June 2022 low, about 44% of Bitcoin addresses were holding at a loss (Glassnode). At the 2022 low, about 54% of Ethereum addresses were (IntoTheBlock). Addresses aren't people: one person can hold many, and an exchange can hold millions of customers in a few.
- In the bear markets after the 2017 and 2021 tops, Bitcoin sellers locked in losses equal to 47.9% and 46.8% of all the profit locked in during the bull market before (Glassnode, Week 50 2022). In dollars, close to half of what was banked on the way up was lost by other sellers on the way down, many of them people who bought near the top.
Many end up selling at a loss
- In a November 2022 survey of 1,441 US adults (about 400 of whom had held crypto), 38% of those holders said they sold it for less than they paid. 28% sold at a profit (LendingTree).
- The Bank for International Settlements estimated that 73% to 81% of retail investors who used crypto apps between 2015 and 2022 likely lost money on their initial investment, largely because new users pile in after prices have already risen (BIS Working Paper 1049, November 2022). That figure comes from a model built on app downloads, nobody counted real account balances to get it, and it's the most pessimistic number we found.
Altcoins fall back much further
- Of the 100 largest altcoins in November 2021, almost two-thirds have lost more than 90% of their value, and only 7 are worth more today than at that peak (CoinMarketCap research, reported by Fortune, 6 November 2026).
- Bitcoin itself peaked near $126,000 in October 2025 and traded around $62,000 in mid-November 2026, about 51% lower (Glassnode; Decrypt).
What the evidence does not show
Whether the average crypto investor is up or down depends on when you ask. Surveys taken near tops find most holders up on paper: in the UK's 2025 survey, taken about six weeks before the October 2025 peak, 62% reported a gain. Surveys taken near bottoms find roughly half underwater and a large share who already sold at a loss. So the claim "most crypto investors never make a profit" goes further than the data. The surveys do show that far more holders sit on paper gains than sell them, and many of those gains are gone by the next low. A selling schedule is built for that problem.
One more finding cuts the other way. Classic research on stock investors found people tend to sell winners too early (Shefrin and Statman 1985; Odean 1998). In crypto the studies are mixed. One finds the same habit in Bitcoin, strongest from 2017 on (Schatzmann and Haslhofer, Digital Finance, 2023). An older, smaller study of the Mt. Gox exchange from 2011 to 2013 finds it reversed during bull runs (Haryanto, Subroto and Ulpah, 2020). The pattern the surveys show most clearly is people holding what they bought near the top all the way through the fall.
4. How we ran the study
This section explains what data we used, where it came from, how we cleaned it, and the math behind every number in section 5. Appendix A has the formulas written out in full.
4.1 The question
We compared these ways of holding the same subnet tokens over the same days, at the same prices:
- No staking: buy the tokens and hold them.
- Stacking: stake the tokens and leave every reward staked.
- Selling daily to USDC: stake the tokens, sell each day's rewards into TAO, and sell that TAO for USDC the same day. The USDC earns 3.75% a year.
- Selling to TAO and keeping the TAO: the same as 3 without the last step. It is shown only to make clear why the last step matters.
For each one we measured the gain or loss in dollars at the end, the worst drop along the way, and, in the exit test, the result on the day an investor actually sells. Every result is per $10,000 put in on the first day.
4.2 The data and where it came from
| Data | What it is | Where it came from | Period |
|---|---|---|---|
| Subnet prices | Each subnet token's daily closing price in TAO, and the chain block each day's close ended on | TaoMarketCap's public data service (api.taomarketcap.com), daily price candles for all 128 subnets | 370 days to 20 November 2026 |
| Price on the last day | Each subnet's price in TAO at block 9,546,026 | The same service, read at the same block for every subnet | 20 November 2026 |
| What a staked token earned | The value of one staked share with a subnet's best validator, read from the chain's own records | The Bittensor chain, through the Opentensor Foundation's public archive node (archive.chain.opentensor.ai) | Every 15 days, 0 to 180 days back from block 9,546,033 |
| Which validator, and today's yearly rate | Every validator's 1-day, 7-day and 30-day yearly rate on every subnet | Taostats validator figures, as stored by taoemissions.com | 19 November 2026 |
| TAO's dollar price | TAO in US dollars at the real time of each day's closing block | The TAO dollar price carried in the same daily candles, matched to each block's own timestamp read from the chain | 365 days to 20 November 2026 |
| USDC interest | 3.75% a year, compounded daily | My own Coinbase USDC rewards rate, November 2026 | Held flat |
| Model portfolios | The Conservative, Mixed and Risk-on members and weights | taoemissions.com subnet score, version 4 | 20 November 2026 |
| Investor behavior | Surveys, on-chain studies and academic papers | Listed in Appendix B | 2015 to 2026 |
How the staking earnings were read. For each subnet we took one validator: the highest-paying one with no warning signs (such as a very small or brand-new stake). At each reading we asked the chain for two stored values on that validator's key: the total subnet tokens staked with it (TotalHotkeyAlpha) and the total shares stakers hold in it (TotalHotkeySharesV2). Tokens divided by shares is what one staked share is worth. Rewards add tokens and never add shares, so the growth of that figure between two readings is exactly what a staker earned after the validator's cut. These are the same two values taoemissions.com reads to show a wallet's stake.
How TAO's dollar price was timed. Each daily candle carries TAO's dollar price at midnight UTC, but the block that closes that candle comes 27 to 34 hours later. We read each closing block's real time from the chain's own clock and drew a straight line between the midnight prices on either side of it. We then checked four dates against an independent hourly TAO price series from a public exchange data service. All four were within 1.05%. Over the study's windows TAO fell from $227.06 to $187.20 (3 months, down 17.6%) and from $278.09 to $187.20 (6 months, down 32.7%).
4.3 Which subnets were included
| 90-day window | 180-day window | |
|---|---|---|
| Subnets on the network on 20 November 2026 | 128 | 128 |
| Too young to have a price on the first day (left out) | 18 | 21 |
| Earnings read from the chain | 102 | 95 |
| Chain reading unusable, so today's yearly rate was held flat | 6 | 10 |
| No reliable earnings figure at all (left out) | 2 | 2 |
A chain reading counted as unusable when the validator's key held no stake at a reading, held under 1,000 tokens, or the value of a share fell between two readings, which rewards alone cannot cause. The random portfolios use only the 95 subnets whose earnings were read from the chain for both windows, so none of their results rest on a held-flat rate.
4.4 The portfolios tested
- Model portfolios: Conservative, Mixed and Risk-on, at the weights published on taoemissions.com on 20 November 2026. These were chosen partly on recent price strength, which flatters any back test of them.
- 10 random portfolios of 4 to 8 subnets, and 90 random portfolios of exactly 10 subnets. Each was drawn by a seeded random number generator: subnets picked without repeats from the 95, each given a random weight between 0.2 and 1.2 and then scaled to add up to 100%. The fixed seed means the same portfolios come out on every run, so anyone re-running the study gets the same draws.
- Every subnet on its own, for the subnet-by-subnet counts.
4.5 The math, in words
- Buying. On day 0 the investor turns $10,000 into TAO and then into the subnet's tokens at that day's prices.
- The daily earning rate. Between two chain readings 15 days apart, the share value grew by some factor. We spread that growth evenly across the days between them as a compounding daily rate (the 15th root of the growth, minus 1). On a day after the last reading, the last stretch's rate carries on.
- No staking. The dollar value on any day is the tokens × that day's subnet price in TAO × that day's TAO dollar price.
- Stacking. Each day the token count grows by that day's earning rate, and the dollar value is the grown token count × the subnet price × the TAO dollar price.
- Selling daily to USDC. Each day's new tokens (the original tokens × that day's rate) are sold at that day's closing price into TAO, and that TAO is sold at that day's TAO dollar price for USDC. The USDC balance earns one day's interest (3.75% a year, compounded daily) before each new deposit. The dollar value on any day is the original tokens at that day's dollar price, plus the USDC balance.
- Selling to TAO and keeping the TAO. The same daily sales into TAO, but the TAO is kept and valued at each day's TAO dollar price.
- Portfolios. A portfolio's dollar value each day is the weighted sum of its members' values. A member with no price on the first day is left out and the rest are scaled up, and if more than half the weight is missing, no result is given.
- The worst drop. For each day we take the highest dollar value so far and measure how far the value sits below it, in percent. The worst drop is the largest of those.
- Share of the loss paid back. When a portfolio lost dollars on price, this is the USDC banked by daily sales divided by that dollar loss.
- Medians. Each median is taken across the portfolios separately for each strategy, so the median for stacking and the median for selling may come from different portfolios.
4.6 The made-up price paths
Real subnet tokens only have about 21 months of history, which is not a full bull and bear cycle. To test a full cycle we built made-up paths for a subnet token's dollar price and ran them at each model portfolio's real average earning rate from the last 180 days (Conservative 42%, Mixed 43% and Risk-on 57% a year), held flat. Because daily sales go straight to dollars, only the token's dollar price matters in these tests.
- The full cycle: the dollar price rises at a steady percentage each day to 5 times its start over 365 days, then falls at a steady percentage each day to 15% of its top over the next 365 days (85% below the top). A second, milder version goes up 3 times and down 70%.
- The bull run nobody sells: over 180 days the dollar price rises in a straight line to double by day 90, then falls in a straight line to end at the start, 20% below it or 40% below it.
4.7 The exit test
To judge the strategies where real investors sell, we needed the day each kind of investor exits.
- The trigger is the dollar value of the tokens alone, which is what a holder watches. The top is that value's highest day.
- An investor in the "within 20% of the top" group sells on the first day after the top when it is 10% below the top, the middle of that band. The "20% to 50%" group sells at 35% below the top, and the "50% to 80%" group at 65% below. The "never sold" group is valued at the lowest day after the top. In the real 90 and 180 day windows, a portfolio that never fell far enough for a group is valued on the last day.
- Both strategies sell their whole position on that same day, so the only difference between them is what happened to the emissions.
- Each group's result is multiplied by its share of investors from the exit mix and added up. No study gives that split for people directly, so we built three mixes (low, realistic and high discipline) from the surveys in section 3, and every result is shown for all three.
4.8 How we checked the work
- The selling-to-TAO-and-keeping-it case reproduces the live back test on taoemissions.com exactly. For example, Conservative over 3 months is +55.16 TAO per 100 TAO, which is +$2,792 per $10,000 at today's TAO price.
- One subnet (Almanac, SN41, over 90 days) was worked out by hand in TAO and matched the code to the cent.
- We re-read the chain every day, instead of every 15 days, for four subnets, including the two with the most volatile rewards. The totals moved by 0.3% or less, so the 15-day readings are accurate enough.
- TAO's dollar price matched an independent hourly price series within 1.05% on all four dates checked.
- Every outside figure in section 3 was re-opened at its source, and the ones marked "checked" in Appendix B were confirmed word for word.
- An independent reviewer read an earlier full draft against the data and the sources, and every claim it found stronger than its source was corrected.
4.9 Limits of this study
- Survivorship. Only subnets still on the network on 20 November 2026 are included. A subnet removed earlier, probably after a bad run, isn't in the data, which may flatter all the strategies.
- Short history, falling market. The real windows are 90 and 180 days, and both were falling markets for subnet tokens in dollars, because TAO itself fell 17.6% and 32.7%. The full-cycle results come from made-up price paths.
- One validator per subnet, the one paying best today. A staker with a different validator earned a different rate.
- Even earnings within each 15-day stretch. Tested and found to move totals by 0.3% or less.
- Same-day conversion. We assume each day's TAO is sold for USDC the same day. In practice many people move TAO to an exchange every few days or weekly, which adds a few days of TAO price risk.
- Costs left out. Fees and price impact fall mostly on the daily seller, who sells twice a day (alpha to TAO, TAO to USDC). Section 5 estimates them.
- One interest rate. 3.75% is my Coinbase rate in November 2026. Rates vary by account and change over time.
- The exit mix is our estimate, built from surveys that measure different things at different times.
- This is a comparison of strategies on past data. It doesn't show what will happen next, and it doesn't prove that daily selling changes how people behave.
5. What we found
All figures are per $10,000 put in. Over these windows TAO itself fell 17.6% (3 months) and 32.7% (6 months), so most results are losses in dollars. The question is which way of handling the emissions lost the least, or made the most.
Finding 1: When prices rise and you sell at the top, stacking wins
Our three model portfolios rose in TAO terms over both windows. Over three months, selling daily to USDC came out slightly ahead in all three. Over six months, stacking came out ahead in all three.
| Portfolio | 3 months: no staking | 3 months: stack | 3 months: sell daily to USDC | 6 months: no staking | 6 months: stack | 6 months: sell daily to USDC |
|---|---|---|---|---|---|---|
| Conservative | +$1,882 | +$3,068 | +$3,107 | +$201 | +$2,563 | +$2,113 |
| Mixed | +$558 | +$1,793 | +$2,040 | -$1,239 | +$773 | +$676 |
| Risk-on | +$1,759 | +$3,336 | +$3,460 | +$419 | +$4,334 | +$3,494 |
This is the "look what I'd make if I held" argument, and on these numbers it's real. It assumes the stacker cashes out on the last day, at a good price.
Finding 2: Across random portfolios, selling daily to USDC usually wins
Over the last six months most subnet tokens lost value in dollars: 87 of 95 subnets, once TAO's own fall is counted. Across the 90 random 10-subnet portfolios:
| 3 months | 6 months | |
|---|---|---|
| Portfolios that lost money on price alone | 87 of 90 | 87 of 90 |
| Median result, no staking | -$2,991 | -$4,990 |
| Median result, stacking | -$2,074 | -$3,538 |
| Median result, selling to TAO and keeping the TAO | -$2,066 | -$3,393 |
| Median result, selling daily to USDC | -$1,726 | -$3,027 |
| Selling daily to USDC beat stacking | 90 of 90 | 80 of 90 |
| Median edge of selling to USDC over stacking | +$343 | +$465 |
| Losing portfolios turned into a gain by daily sales | 2 | 8 |
| Median dollars banked in USDC | $1,200 | $1,990 |
| Median share of the price loss paid back by daily sales | 41% | 39% |
Subnet by subnet, selling daily to USDC beat stacking for 97 of 102 subnets over three months and 84 of 95 over six.
Why the second step matters. Selling into TAO and keeping the TAO did little better than stacking: its median was about the same over three months and $145 better over six, and it beat stacking in only 58 and 64 of the 90 portfolios. Selling all the way to USDC beat keeping the TAO in all 90 portfolios in both windows, because TAO fell 32.7% over the six months. TAO kept as TAO falls with TAO, and only the dollars held their value.
Costs. Fees and price impact fall almost entirely on the daily seller, who sells twice a day (alpha to TAO, then TAO to USDC) while a stacker sells once. We re-ran the six-month test with a cost on every daily sale. At 0.5% per sale, the median edge over stacking was +$456 and selling to USDC still won in 80 of 90. At 1% per sale it was +$446, again 80 of 90. Costs barely move it because they apply only to the emissions sold, about $2,000 per $10,000 over six months.
Interest. Over six months the 3.75% added only about $18 to the median result, because the USDC balance builds up a little each day. The interest grows in importance the longer you run the plan and the more you have banked.
Finding 3: Selling daily to USDC cuts the worst drop
The worst drop is the number that makes people panic and sell at the bottom.
| Median worst drop, in dollars | No staking | Stacking | Selling daily to USDC |
|---|---|---|---|
| 90 random portfolios, 3 months | 47.0% | 43.6% | 40.3% |
| 90 random portfolios, 6 months | 54.3% | 46.2% | 39.9% |
Selling daily to USDC had the smallest worst drop of the three in every one of the 90 random portfolios, in all three model portfolios and in every single subnet tested, in both windows.
Finding 4: Stackers give their gains back
Over six months, the median stacker's best moment was +$1,652 per $10,000, and the median finish was -$3,538. 89 of the 90 stackers handed back half or more of their best gain by the end. Over three months it was 90 of 90, with a median best of +$3,666 and a median finish of -$2,074.
Finding 5: In a bull run nobody sells, selling daily to USDC wins clearly
We ran a path where the token's dollar price doubles by the halfway point and then falls back, at the Conservative portfolio's real earning rate, over six months.
| Where the price ends | No staking | Stack, never sold | Sell daily to USDC | Of which already banked in USDC |
|---|---|---|---|---|
| Back where it started | $0 | +$2,308 | +$3,145 | $3,145 |
| 20% below the start | -$2,000 | -$154 | +$1,040 | $3,040 |
| 40% below the start | -$4,000 | -$2,615 | -$1,066 | $2,934 |
At the top of that run, the stacker was up $12,188. If they didn't sell, most of it was gone by the end. The daily seller banked part of every day's pay in dollars at the high prices on the way up, and those dollars stay banked wherever the token goes next.
Finding 6: Judged at the moments real people sell, selling daily to USDC usually comes out ahead
This is the test that matters most. We built a full two-year cycle: the token's dollar price rises five times over a year, then falls 85% from its top over the next year, about the fall my own portfolio took. Both strategies sell their whole position at the same moment, at each of the points where real investors exit. These figures are for the Conservative portfolio, which earned 42% a year, held flat for both years.
| When the investor finally sells | Price vs the top | Stack | Sell daily to USDC |
|---|---|---|---|
| Within 20% of the top | -10% | +$59,978 | +$46,638 |
| 20% to 50% below the top | -35% | +$44,459 | +$37,106 |
| 50% to 80% below the top | -65% | +$23,657 | +$25,642 |
| Never sold, rode it to the low | -85% | +$7,410 | +$18,180 |
In this test stacking won for investors who sold within 35% of the top and lost for those who sold 65% or more below it. For the Risk-on portfolio, with its higher earning rate, stacking was still slightly ahead at 65% below the top. For the half of investors who never sell, selling daily to USDC finished at $18,180 against $7,410.
Then we weighted those outcomes by how many investors exit at each point. No study gives that split directly, so we built three versions from the surveys in section 3.
| Where investors finally sell | Low discipline | Realistic | High discipline |
|---|---|---|---|
| Within 20% of the top | 10% | 12% | 20% |
| 20% to 50% below the top | 10% | 13% | 15% |
| 50% to 80% below the top | 26% | 25% | 25% |
| Never sold, rode it to the low | 54% | 50% | 40% |
The realistic mix leans on the SEC's 11% who "sold and made a lot," the 54% to 55% who held through the 2022 crash, the 26% who sold a lot or all of their crypto in mid-2022 (Civic Science), and the 38% who said they sold at a loss (LendingTree). We value the never-sold group at the low. That is the point where the comparison favors daily selling most, and if those holders keep holding and the price recovers, the gap narrows.
This table shows selling daily to USDC minus stacking, per $10,000, for the low, realistic and high discipline mixes:
| Portfolio | Up 5x, then down 85% | Up 3x, then down 70% |
|---|---|---|
| Conservative | +$4,262 / +$3,324 / +$1,033 | +$1,603 / +$1,192 / +$209 |
| Mixed | +$4,239 / +$3,281 / +$942 | +$1,525 / +$1,107 / +$108 |
| Risk-on | +$3,013 / +$1,723 / -$1,424 | -$652 / -$1,169 / -$2,399 |
Under the realistic mix, selling daily to USDC came out ahead in five of the six cycle cases. Stacking won for Risk-on in the milder cycle, because a 57% yearly rate compounds hard and a 70% fall leaves more of it standing. On the real data with the realistic mix, selling daily to USDC beat stacking in all 90 random portfolios over three months (median edge +$290) and in 79 of 90 over six (median edge +$359).
Finding 7: Selling daily to USDC roughly halves the cost of bad timing
The gap between an investor who sells at a good moment and one who sells at the worst moment is a clean way to measure risk. In the up 5x, down 85% cycle, that gap was $52,568 for a Conservative stacker and $28,458 for a Conservative daily seller, 46% narrower. For Risk-on it was $58,413 against $25,344, 57% narrower. The gap was narrower for the daily seller in every cycle and every portfolio we modelled.
Part of that is upside given up: the Conservative daily seller's best exit was $13,340 below the stacker's, and its worst exit was $10,769 above.
6. Pros and cons of each strategy
Stacking emissions
Pros
- It makes the most money in a market that rises and is still up when you sell.
- Your token count compounds, so every price gain applies to a bigger pile.
- There is nothing to manage day to day.
Cons
- Every reward stays exposed to both the subnet token's price and TAO's price, so a fall in either hits your earnings as hard as your principal.
- It beats daily selling only if you sell reasonably close to the top. In the SEC survey, about 1 current owner in 9 said they "sold and made a lot of money."
- The worst drops are bigger, and that is when people panic and sell at the worst time.
- Your gains stay on paper until you act, and the research says most people don't act.
Selling emissions daily to USDC
Pros
- It banks a slice of every day's pay in dollars, at that day's price, with no timing call.
- The dollars are out of the reach of both the subnet token's fall and TAO's fall, and they earn interest while they sit.
- The worst drop was the smallest in every portfolio and every subnet we tested.
- It came out ahead in most of the cases we modelled once real exit behavior was counted.
- It turns paper gains into money you can spend, which makes it a real second income.
- I think it builds a habit of taking profits. My team found no study that measures this directly, so treat it as my opinion.
Cons
- It earns less than stacking when prices rise and you do sell near the top.
- It takes two steps each time, one from your wallet and one on an exchange, and each has a small cost.
- The USDC sits on an exchange, so you take on that exchange's risk and the stablecoin's.
- Each sale is one more transaction to report for tax, and the rewards themselves may already count as income when you receive them, depending on where you live.
7. The "but if I hold and sell at the top" argument
Stacking's best numbers in this study all come from one scenario: you hold every reward through the run-up and sell near the peak. Our model portfolios over six months are that case, and stacking won there.
The problem is the assumption. The SEC survey found that 37.1% of current crypto owners watched their holdings gain a lot without selling, while 11.0% said they sold and made a lot. In each of Bitcoin's last two bear markets, sellers locked in losses close to half the size of the profits banked in the bull market before. Almost two-thirds of 2021's top altcoins are down more than 90%. And then there's me, with $2.6 million on the screen and no plan to sell.
Almost nobody pulls off a perfect exit. Judged against the exits people actually make, selling daily to USDC came out ahead in most of the cases we modelled, though stacking still won for the Risk-on portfolio in the milder cycle.
8. What selling emissions daily will not do
- It won't protect your principal. The tokens you staked still ride the subnet's price and TAO's price. Only the daily pay is moved into dollars.
- It won't make a falling token a good investment. Over six months 87 of 95 subnet tokens lost value in dollars. Daily sales paid back a median 39% of the price loss for the 10-subnet portfolios, and the median portfolio still finished down $3,027 per $10,000.
- It doesn't remove the risks of the subnets themselves. Subnets can be removed from the network, change owners, cut their rewards, or trade too thinly to sell a large amount without moving the price. Validators can change their fees, and Bittensor's own rules change with network upgrades.
- It adds exchange and stablecoin risk. The dollars sit on an exchange in USDC, and interest rates change.
- Past results are past results. The last six months were a falling market for subnet tokens and for TAO. In a long rising market where you do sell near the top, stacking would usually come out ahead.
Nothing in this paper is investment advice or tax advice.
9. How I think the average investor should approach crypto, and Bittensor
This is what I'd tell a friend after doing this study, and after watching my own portfolio fall 85%.
Have a selling plan before you buy, and make it automatic. In the SEC survey, more than three times as many owners held through a big gain (37.1%) as sold and made a lot (11.0%). A lab experiment found that automatic sell orders cut the habit of holding onto losing positions, while reminders alone did not (Fischbacher, Hoffmann and Schudy, Review of Financial Studies, 2017). The orders didn't make people sell more of their winners, and no study has tested daily selling of staking rewards, so this is where my experience comes in.
Own assets that pay you to hold them. Most crypto pays nothing to hold, so when it falls back you're left with only the fall. Over six months, our 90 random subnet portfolios lost a median $4,990 per $10,000 on price alone, but $3,027 once the daily pay was sold for dollars. The tokens and prices were the same in both cases, and the only difference was collecting the pay.
Bank the pay every day, in dollars. Sell the emissions into TAO, and sell the TAO for USDC. Stopping at TAO leaves you exposed to TAO's own crash: over these six months TAO fell 32.7%, and keeping the TAO lost more than taking dollars in every one of the 90 portfolios. The USDC earns interest while it sits, and it's real income you can use.
Spread it across several subnets. Over six months, random 10-subnet portfolios ranged from -$4,963 to +$5,492 per $10,000 selling daily to USDC, so a single subnet can decide your year. Our model portfolios screen for revenue, buybacks, owner lockups and price stability, and they're a reasonable starting point for your own research.
Size it so a crash doesn't break you. Assume an 80% fall is possible, because it has happened over and over in crypto. Put in what you could watch fall that far without selling at the bottom.
The case for moving idle tokens into Bittensor
A lot of people are sitting on tokens that pay them nothing, holding for the next run with no plan to sell it. The research says most of them will ride it back down.
For someone who is going to hold crypto either way, I think there's a strong argument for moving some of those idle funds into TAO and subnet staking with a daily selling plan. Most subnets with a named best validator pay 30% a year or more. Selling that pay every day and taking it to USDC turns it into a daily flow of dollars that keeps arriving whether the market is up or down, though the amount shrinks when the subnet token falls or rewards are cut. It's a second income, it earns interest once it's banked, it cushions the losses on your principal, and it means you don't have to nail the top to come out with something.
That move has real risks, and they belong in the decision. You'd be concentrating in one network, so TAO's own price becomes your biggest risk on the principal. Subnet tokens are among the most volatile assets in crypto, rewards change as the network changes, and trading one token for another can be a taxable sale. Do your own research, start with a size you can live with, and spread it across several subnets.
My view after this study is that stacking only beats daily selling if you sell reasonably close to the top, and the surveys say most people don't. Selling emissions daily and taking them to dollars gives up some upside, and in return the plan still works when you get the timing wrong.
10. Conclusion
For the way most people actually invest, selling your emissions every day and taking them to dollars beats stacking. This study supports that conclusion without hedging:
- Selling daily to USDC beat stacking in all 90 random portfolios over three months and in 80 of 90 over six.
- Under realistic investor behavior it came out ahead in 79 of 90 real portfolios over six months and in five of the six full cycles we modelled.
- It had the smallest worst drop in every portfolio and every subnet we tested.
- Stopping halfway and keeping the TAO lost more than taking dollars in all 90 portfolios. The dollars are what protect you.
Stacking wins in one situation: you hold through the run-up and sell at or near the top. Our model portfolios over six months are that case, and so is the Risk-on portfolio in the milder cycle. Investors who time it that well do make more by stacking. They also carry much bigger losses whenever the timing goes wrong, and most of the time it goes wrong. In the SEC survey, about 1 current owner in 9 said they sold and made a lot of money, and in our realistic mix half of investors never sell at all.
The trade is lopsided. Selling daily limits your upside somewhat, and it limits your downside far more, for the investors most likely to need it:
| Best outcome: stack | Best outcome: sell to USDC | Worst outcome: stack | Worst outcome: sell to USDC | |
|---|---|---|---|---|
| 90 real portfolios, 3 months | +$1,735 | +$1,925 | -$4,025 | -$3,595 |
| 90 real portfolios, 6 months | +$5,608 | +$5,492 | -$5,685 | -$4,963 |
| Full cycle, Conservative (5x up, 85% down) | +$59,978 | +$46,638 | +$7,410 | +$18,180 |
| Full cycle, Risk-on (5x up, 85% down) | +$71,805 | +$50,782 | +$13,392 | +$25,437 |
- On the real data, the trade barely cost anything. Over six months the best portfolio gave up $117 of its gain (2%) by selling daily to USDC, while the worst portfolio lost $722 less. Over three months selling daily to USDC came out ahead at both ends.
- In the full cycles, the investor who sells near the top gives up 21% to 29% of the upside by selling daily. The investor who never sells ends with 1.1 to 2.5 times as much by selling daily, and that's where half of investors end up.
- Along the way, the median worst drop over six months was 46.2% stacking and 39.9% selling daily to USDC, and stacking's swings were bigger in every portfolio we tested.
The extra upside from stacking goes to the few who time the top, and the extra downside lands on everyone else.
Even for the investor who plans to sell near the top, stacking is a bad trade. The fair way to judge it is to compare the extra money stacking can make you with the extra risk it makes you carry. Here the risk is how far your result can swing between selling at the best moment and selling at the worst one. In every case we tested, the extra risk was bigger than the extra reward:
| Case, per $10,000 | Extra upside from stacking (best exit) | Extra swing from stacking (best exit minus worst exit) | Risk per $1 of extra upside |
|---|---|---|---|
| 90 real portfolios, 6 months | $117 | $838 | $7.16 |
| Conservative, up 5x then down 85% | $13,340 | $24,110 | $1.81 |
| Conservative, up 3x then down 70% | $6,806 | $10,505 | $1.54 |
| Mixed, up 5x then down 85% | $13,727 | $24,613 | $1.79 |
| Mixed, up 3x then down 70% | $7,019 | $10,675 | $1.52 |
| Risk-on, up 5x then down 85% | $21,023 | $33,069 | $1.57 |
| Risk-on, up 3x then down 70% | $11,126 | $13,119 | $1.18 |
Over three months on the real portfolios there was no extra upside at all: selling daily to USDC came out ahead at the best end and at the worst end.
So for every extra dollar stacking can make a perfect timer, it adds between $1.18 and $1.81 of extra swing in the full cycles, and more than $7 on the real six months of data. The plan to "stack and sell near the top" only pays if both your timing and your thesis are right. If the crash comes before you sell, or the project you believed in never gets back to its high, you take the whole swing. That is what happened to my TAO in 2024. A trade that risks more than it can make doesn't make sense, even for a skilled investor.
My conclusion: for the average crypto investor, and for anyone staking in Bittensor, selling your emissions every day and taking them to dollars is the better strategy. Stacking is a bet on your own ability to sell near the top, and the evidence says most people lose that bet. The limits in section 8 still apply: your principal still rides the market, and past results don't promise future ones.
11. What taoemissions.com does for you
Everything in this paper comes down to one habit: sell the pay every day and take it to dollars. Taoemissions.com is built to make the first half automatic and the whole plan visible.
- See every emission you earn. Add your wallet's public address (never a seed phrase or private key) and the site counts your emissions per subnet, in TAO and in dollars, with a running tax estimate at the rate you set.
- Know when to sell. When a subnet's fresh emissions reach 0.1 TAO, the dashboard recommends selling them, so the schedule runs without you watching charts.
- Sell from your own wallet (Pro). You click Sell emissions, your own browser wallet opens, and you approve the sale yourself. Only the emissions are sold, never your principal, and the site never holds your keys. The TAO lands in your wallet, and you send it to your exchange and sell it for USDC.
- Stake with the validator that pays best (Pro). Your validator's 7-day return sits next to the best one on the same subnet, and Pro names it.
- Build a portfolio and back test it (Pro). Three model portfolios, a builder for your own mix, and a back test of the last 3 and 6 months.
What it costs against what it earns. Pro is $19.99 a month. A 100 TAO stake earning 40% a year pays about 3.3 TAO a month in emissions, about $624 a month at today's TAO price of about $187, so Pro costs about 3% of what that stake earns. In this study, daily sales taken to USDC paid back a median 39% of the price loss on 10-subnet portfolios over six months. Basic is $9.99 a month, and the first 14 days are free with no card needed.
Appendix A: The formulas
Notation, for one subnet over a window of n days (day 0 is the buying day, day n is 20 November 2026):
- p(t) is the subnet token's closing price in TAO on day t.
- u(t) is TAO's price in dollars at the real time of day t's closing block.
- r(t) is the staking rate earned on day t, as a fraction (0.001 means 0.1% that day).
- N = 10,000 ÷ [p(0) × u(0)] is the number of tokens bought with $10,000.
- y = 1.0375^(1/365) − 1 is one day's USDC interest at 3.75% a year.
The daily earning rate from the chain. At a chain reading, s = TotalHotkeyAlpha ÷ TotalHotkeySharesV2 is the value of one staked share. Between two readings d days apart, with share values s₁ and s₂:
- growth g = s₂ ÷ s₁
- daily rate r = g^(1/d) − 1, the same on every day of that stretch
Over a whole window the daily rates compound back exactly to the growth read from the chain. The code checks this on every subnet read from the chain, to within one part in a million.
Value on day t, in dollars:
- No staking: V(t) = N × p(t) × u(t)
- Stacking: V(t) = N × (1 + r(1)) × … × (1 + r(t)) × p(t) × u(t)
- Selling daily to USDC: V(t) = N × p(t) × u(t) + D(t), where D(t) = D(t − 1) × (1 + y) + N × r(t) × p(t) × u(t) × (1 − c), with D(0) = 0. Here c is the cost of a day's two sales: 0 in the main results, and 0.5% or 1% in the cost check.
- Selling to TAO and keeping the TAO: V(t) = N × p(t) × u(t) + T(t) × u(t), where T(t) = N × [r(1) × p(1) + … + r(t) × p(t)] is the TAO banked by day t
Result: V(n) − 10,000, in dollars per $10,000 put in.
Portfolio: with weights w₁ … wₖ that add up to 1, the portfolio value is V(t) = w₁ × V₁(t) + … + wₖ × Vₖ(t).
Worst drop: for each day t, M(t) is the highest value on any day up to t. The worst drop is the largest value of [M(t) − V(t)] ÷ M(t), shown as a percent.
Share of the price loss paid back: D(n) ÷ [10,000 − N × p(n) × u(n)], when N × p(n) × u(n) is below 10,000.
Exit trigger for the exit test: the dollar value of the tokens alone, H(t) = Σ wᵢ × Nᵢ × pᵢ(t) × uᵢ(t). The top T is the day H(t) is highest. An investor in a group with depth z (0.10, 0.35 or 0.65) exits on the first day t after T with H(t) ≤ (1 − z) × H(T). The never-sold group is valued on the day after T with the lowest H(t).
Weighted exit result: the sum over the four groups of (that group's share of investors × V on that group's exit day − 10,000).
Made-up full cycle (365 days up, 365 days down): the token's dollar price is P(t) = 5^(t/365) for t up to 365, then P(t) = 5 × 0.15^((t − 365)/365). The milder version uses 3 and 0.30. In these tests u(t) = 1 and p(t) = P(t), so the price is in dollars directly. The daily earning rate is held flat at r = the weighted average of each member's daily rate, g^(1/180) − 1, where g is that member's 180-day growth read from the chain.
Bull run nobody sells (180 days): P(t) = 1 + t/90 up to day 90, then a straight line from 2 down to the ending price (1.0, 0.8 or 0.6) at day 180.
Random portfolios: a seeded random number generator (Mulberry32) picks subnets without repeats from the 95 eligible ones. Each gets a weight of 0.2 + a random number between 0 and 1, the weights are scaled to add up to 100%, and they are rounded to 0.1%. Seed 20261121 gives the 10 portfolios of 4 to 8 subnets, and seed 90210121 gives the 90 portfolios of 10.
The code: every figure was produced by scripts that read the saved data files listed in section 4.2. Nothing was typed in by hand, so the study can be re-run on any later date and checked against this paper.
Appendix B: Sources
"Checked" means the original was re-opened and the figure confirmed on 21 November 2026.
- US SEC Office of the Investor Advocate, Carman, Chin, Cook and Zimmerman, "Experiences with Crypto Assets in the U.S. Population: Findings from THRIVE," OIAD Working Paper 2026 No 4, July 2026. Listed on the SEC's Investor Research page under Working Papers. https://www.sec.gov/files/experiences-crypto-assets-us-population.pdf (checked).
- LendingTree, "Tales From the Crypto Crypt: 38% of Investors Have Lost More Money Than Made It," 23 January 2023 (1,441 US adults, 15 to 21 November 2022). https://www.lendingtree.com/debt-consolidation/crypto-investors-survey/ (checked).
- Fortune (A. Chernikova), "Nearly two-thirds of 2021's top 100 altcoins have lost more than 90%, CoinMarketCap data shows," 6 November 2026. https://fortune.com/2026/11/06/altcoins-lost-90-percent-coinmarketcap-data/ (checked).
- Bank for International Settlements, Auer, Cornelli, Doerr, Frost and Gambacorta, "Crypto trading and Bitcoin prices: evidence from a new database of retail adoption," BIS Working Paper 1049, 14 November 2022. https://www.bis.org/publ/work1049.htm (checked, summary page).
- Glassnode, The Week On-chain, Week 50 2022. https://research.glassnode.com/the-week-onchain-week-50-2022/ (checked).
- Glassnode, The Week On-chain, Week 47 2021. https://research.glassnode.com/the-week-onchain-week-47-2021/ (checked).
- Kogan, Makarov, Niessner and Schoar, "Are Cryptos Different? Evidence from Retail Trading," NBER Working Paper 31317; Journal of Financial Economics vol. 159, 2024. https://www.nber.org/papers/w31317 (checked, abstract).
- UK Financial Conduct Authority and YouGov, Cryptoasset consumer research, 2021, 2024 (wave 5) and 2025 (wave 6). https://www.fca.org.uk/publications/research/research-note-cryptoasset-consumer-research-2021, https://fca.org.uk/publication/research-notes/cryptoasset-consumer-research-2024-wave-5.pdf and https://www.fca.org.uk/publication/research-notes/cryptoasset-consumer-research-2025-wave-6.pdf
- Civic Science survey, mid-2022, reported by IBTimes, 1 August 2022. https://ibtimes.com/wealthy-crypto-owners-weather-crash-while-low-income-investors-sell-holdings-survey-3592901
- Appinio survey, reported by Cointelegraph, 7 July 2022. https://cointelegraph.com/markets/survey-shows-55-of-crypto-investors-chose-to-hodl-as-bitcoin-and-altcoin-prices-collapsed
- Glassnode, reported by Cointelegraph, Bitcoin addresses in profit, 21 June 2022. https://cointelegraph.com/markets/that-s-not-hodling-over-50-of-bitcoin-addresses-still-in-profit
- IntoTheBlock, reported by Bitcoinist, Ethereum addresses in profit at past lows, 17 September 2024. https://bitcoinist.com/only-61-ethereum-addresses-profit-past-bear-markets
- Glassnode, The Week On-chain, Week 40 2025 (the October 2025 top). https://research.glassnode.com/the-week-onchain-week-40-2025/
- Decrypt, "Bitcoin investors capitulating," 17 November 2026. https://decrypt.co/383466/bitcoin-investors-capitulating-historical-bottom-signal
- Schatzmann and Haslhofer, "Exploring investor behavior in Bitcoin: the disposition effect," Digital Finance, 2023. https://arxiv.org/abs/2010.12415
- Haryanto, Subroto and Ulpah, "Disposition effect and herding behavior in the cryptocurrency market," Journal of Industrial and Business Economics 47(1), 2020. https://ideas.repec.org/a/spr/epolin/v47y2020i1d10.1007_s40812-019-00130-0.html (abstract).
- Fischbacher, Hoffmann and Schudy, "The Causal Effect of Stop-Loss and Take-Gain Orders on the Disposition Effect," Review of Financial Studies 30(6), 2017. https://ideas.repec.org/a/oup/rfinst/v30y2017i6p2110-2129..html
- Shefrin and Statman (1985) and Odean (1998), the classic studies of selling winners too early.
- TaoMarketCap public data service, daily subnet price candles and prices at block 9,546,026. https://api.taomarketcap.com/public/v1
- Opentensor Foundation public archive node for the Bittensor chain: share values (TotalHotkeyAlpha, TotalHotkeySharesV2) and block timestamps. https://archive.chain.opentensor.ai
- Taostats validator figures (1, 7 and 30-day yearly rates per validator), as stored by taoemissions.com on 19 November 2026. https://taostats.io
- Independent hourly TAO dollar prices used only to check our TAO prices: CoinGecko (api.coingecko.com) or Binance (api.binance.com), whichever answered.
- taoemissions.com subnet score version 4 and model portfolios, 20 November 2026.
- CoinGecko, Bittensor (TAO) price page: all-time high of $757.60 on 7 March 2024. https://www.coingecko.com/coins/bittensor (also listed by Decrypt: https://decrypt.co/price/bittensor)
- Coinbase, USDC Rewards overview: the rate varies by country and account type and can change monthly. https://help.coinbase.com/en/coinbase/getting-started/crypto-education/usd-coin-rewards-faq (the 3.75% used here is the author's own account rate, shown as 3.75% APY in his Coinbase account on 21 November 2026, on USDC from emissions he has sold since 11 November).
Start selling your emissions on a schedule
Track every emission, see the best validator on each subnet, and sell your emissions from your own wallet every day. The first 14 days are free and you don't need a card.