This page in one lineEvery month we split the big coins into five groups using one simple rule, then check which group made the most money. We tried 7 rules. The short answer: none of them beat just holding Bitcoin.
Which Ways Of Picking Coins Made Money
Every month we sort up to 80 big coins into five groups by one simple thing about them, then see which group made the most money over the next month. Tested from 2022-07-31 to 2026-09-30. The best sorting rule was “Few buyers so far”: its top group beat its bottom group by +1.4% a month, and it won in 58% of months. The worst was “Busy coins” at -4.0% a month. 3 of the 7 rules made money after costs and won more months than they lost. Be careful with that count: “Calm coins” only clears zero by +0.11% a month, which is too small to build anything on. Here is the blunt version. We set two tests before starting: win more than 60% of months, and make money after costs. No rule passed both. “Few buyers so far” makes the most money but wins 58% of months, under the 60% bar. So treat everything below as a hint, not a result. Now the catch on that +1.4%: it assumes you buy the top group AND bet against the bottom group. Betting against a coin costs extra money that we have not charged here, so the real number is lower. If you only buy the top group, you get what the top bar shows and nothing more.
Which coins are in each group right now?
This is the “Few buyers so far” rule, the one that sorted coins best (none of the rules beat simply holding Bitcoin), applied to the market as it stands on 2026-09-30. Group 1 holds the coins this rule ranks highest and group 5 the ones it ranks lowest. That is a ranking, not a shopping list: read the rest of the page before doing anything with it, because this rule did not beat simply holding Bitcoin. These lists change every month, so treat this as today's answer, not a permanent one.
Drag the chart sideways to see the rest of it
| Group | What put it here | Coins | Which coins |
|---|---|---|---|
| Group 1 (top) | What this rule likes most | 15 | TRON, Zcash, Quant, Official World Liberty Financial, JUST, PancakeSwap, Stacks, Injective Protocol, Dash, Ether.fi, Curve DAO Token, Starknet, Pudgy Penguins, Pendle, Arweave |
| Group 2 | In between | 14 | Bitcoin, Stellar, Arbitrum, Cosmos, Raydium, Virtuals Protocol, Tezos, Conflux Network, Compound, THETA, IOTA, dogwifhat, Trust Wallet Token, Axie Infinity |
| Group 3 | In between | 15 | Solana, Near Protocol, Uniswap, Hedera Hashgraph, Aave, Ethereum Classic, Polygon Ecosystem Token, Algorand, VeChain, OFFICIAL TRUMP, Kaia, Lido DAO, Celestia, Ethereum Name Service, LayerZero |
| Group 4 | In between | 14 | Chainlink, Bitcoin Cash, Ethena, BitTensor, Worldcoin, Internet Computer, Jupiter Exchange Token, Render, Filecoin, Aptos, The Graph, JasmyCoin, THORChain, Convex Finance |
| Group 5 (bottom) | What this rule likes least | 14 | Ethereum, BNB, XRP, Dogecoin, Cardano, Litecoin, Sui, Avalanche, Ondo, Pyth Network, Artificial Superintelligence Alliance, Sei, Optimism, SUN |
What this does not mean. This is not advice to buy these coins. It is what one simple rule picks out, and the rule is only a little better than a 50-50 guess.
How does the test work?
Picture five buckets. Every month-end, the coins trading at the time (up to 80) are lined up by one trait and dealt into the buckets: the strongest fifth in bucket 1, the weakest fifth in bucket 5. Equal money goes into every coin in a bucket. A month later, the buckets are counted and the coins are dealt again. If a rule matters, group 1 should beat group 5, and should do it in most months. Group 1 always means the coins the rule likes most and group 5 the ones it likes least, so that holds for every rule below. A clean staircase down from group 1 to group 5 is the sign of a rule that really separates coins; a jumbled middle means only the two ends are worth reading. Here is the whole market for comparison: Bitcoin and an equal-money mix of every coin. One warning before anything else: these are the coins that are large today, so coins that died along the way are missing, and every line here looks rosier than a real investor’s would have.
Drag the chart sideways to see the rest of it
This is the backdrop. Any trait has to be judged against what simply holding would have done.
Trait: Few buyers so far
This one COUNTS PEOPLE. Of everyone holding a bet on this coin, only a small slice were betting the price goes up, so most of the crowd has not picked it yet. Do not mix this up with “Cheap to bet on a rise” above, which measures a fee instead. The two sound similar and gave opposite answers.
Drag the chart sideways to see the rest of it
Bucket 1 minus bucket 5: +1.6% a month before costs, +1.4% after. Ahead in 58% of 50 months. Bitcoin and Ether are so much bigger than the rest that they can drag the whole answer around, so we also ran it with those two taken out: the gap becomes +1.4%. If that number is close to the main one, the rule is not secretly just a Bitcoin story. On average 59 of the coins could be measured each month.
Trait: Bets piling up
The total pile of money sitting in live bets on this coin grew the most this month. A bigger pile means more people have joined in.
Drag the chart sideways to see the rest of it
Bucket 1 minus bucket 5: +0.9% a month before costs, +0.6% after. Ahead in 51% of 47 months. Bitcoin and Ether are so much bigger than the rest that they can drag the whole answer around, so we also ran it with those two taken out: the gap becomes +0.9%. If that number is close to the main one, the rule is not secretly just a Bitcoin story. On average 63 of the coins could be measured each month.
Trait: Calm coins
Coins whose price wobbled least day to day over the last month.
Drag the chart sideways to see the rest of it
Bucket 1 minus bucket 5: +0.3% a month before costs, +0.1% after. Ahead in 64% of 50 months. Bitcoin and Ether are so much bigger than the rest that they can drag the whole answer around, so we also ran it with those two taken out: the gap becomes -0.1%. If that number is close to the main one, the rule is not secretly just a Bitcoin story. On average 70 of the coins could be measured each month.
Trait: Losers bounce back
Coins that fell most over the last month go in the top bucket.
Drag the chart sideways to see the rest of it
Bucket 1 minus bucket 5: -0.2% a month before costs, -0.5% after. Ahead in 48% of 50 months. Bitcoin and Ether are so much bigger than the rest that they can drag the whole answer around, so we also ran it with those two taken out: the gap becomes -0.7%. If that number is close to the main one, the rule is not secretly just a Bitcoin story. On average 69 of the coins could be measured each month.
Trait: Winners keep winning
Coins that rose most over the last three months, ignoring the most recent week. One caution specific to this rule: it looks back three months but is re-sorted every month, so each month's reading shares about two thirds of its days with the month before. Its months are not fully separate tests, and its 49 months are closer to about 16 independent ones.
Drag the chart sideways to see the rest of it
Bucket 1 minus bucket 5: -1.4% a month before costs, -1.6% after. Ahead in 52% of 50 months. Bitcoin and Ether are so much bigger than the rest that they can drag the whole answer around, so we also ran it with those two taken out: the gap becomes -1.6%. If that number is close to the main one, the rule is not secretly just a Bitcoin story. On average 68 of the coins could be measured each month.
Trait: Cheap to bet on a rise
This one is about a FEE, not about people. On betting sites, whichever side is more crowded pays a small fee to the other side. These are the coins where betting on a rise was cheapest, because that fee was smallest. Do not mix this up with “Few buyers so far” below, which counts people instead. The two sound similar and gave opposite answers.
Drag the chart sideways to see the rest of it
Bucket 1 minus bucket 5: -2.9% a month before costs, -3.1% after. Ahead in 43% of 37 months. Bitcoin and Ether are so much bigger than the rest that they can drag the whole answer around, so we also ran it with those two taken out: the gap becomes -2.9%. If that number is close to the main one, the rule is not secretly just a Bitcoin story. On average 65 of the coins could be measured each month.
Trait: Busy coins
Coins where the most money changed hands over the last month.
Drag the chart sideways to see the rest of it
Bucket 1 minus bucket 5: -3.9% a month before costs, -4.0% after. Ahead in 40% of 50 months. Bitcoin and Ether are so much bigger than the rest that they can drag the whole answer around, so we also ran it with those two taken out: the gap becomes -3.0%. If that number is close to the main one, the rule is not secretly just a Bitcoin story. On average 70 of the coins could be measured each month.
Which traits paid, still paid once you count the cost of buying and selling?
Every time a coin moves into or out of a bucket, it has to be bought or sold. That costs money. The first bar is the raw gap between bucket 1 and bucket 5; the second bar charges 0.10% for every entry and exit. A trait that swaps most of its coins every month loses a lot of its advantage to those costs.
Drag the chart sideways to see the rest of it
Monthly figures. The first bar of each pair is before the cost of buying and selling, the second after. Do not multiply by 12 to get a year, because money compounds: the best rule's +1.40% a month works out to +18% over a year if it kept going. The year-by-year table lower down is also in monthly gaps, so compare it with the +1.40% and not with the yearly figure.
What this does not mean. 0.10% per trade is fair for a small account on a large exchange. Big money moving fast pays more, and thinly traded coins (few buyers and sellers) cost more still.
How often was each trait ahead?
An average can be carried by one huge month. The hit rate cannot: it counts the share of months where bucket 1 beat bucket 5, no matter by how much. Guessing at random gets 50 out of 100.
Drag the chart sideways to see the rest of it
Each bar shows the share of months, then how many months that rule could be tested over. Anything touching the dashed line is no better than guessing. We decided in advance that a rule needs to clear 60% before we would call it real. 1 did: “Calm coins” at 64%.
What this does not mean. A 60% hit rate over 50 months is about 30 wins. That is a hint, not proof: guess at random 50 times and you would get that many heads by pure luck reasonably often.
What did $100 do in the best trait’s buckets?
Follow $100 through bucket 1 of “Few buyers so far”, through its bucket 5, and through plain Bitcoin. Costs are charged on the buckets.
Drag the chart sideways to see the rest of it
Each line is charged its own buying and selling costs, worked out from how much of that bucket is swapped each month: 0.10% a month for group 1 and 0.08% for group 5. Remember that these are today’s large coins, so every line is rosier than a real investor’s would have been; the gap between the two bucket lines is the part to trust.
What this does not mean. Bucket 1 can lose money in a falling market and still be the right bucket. The test is the gap, not the level.
Did the traits work every year, or just once?
The same gap, split by calendar year. A trait that is positive in every year is far more believable than one that made all its money in a single stretch.
Drag the chart sideways to see the rest of it
| Rule | 2022 monthly gap, before costs | 2023 monthly gap, before costs | 2024 monthly gap, before costs | 2025 monthly gap, before costs | 2026 monthly gap, before costs | Months ahead, whole period |
|---|---|---|---|---|---|---|
| Few buyers so far | +4.7% | +1.9% | -2.0% | +3.0% | +2.4% | 58% |
| Bets piling up | +1.5% | +6.3% | -3.4% | +0.4% | +0.1% | 51% |
| Calm coins | +11.5% | -12.2% | +2.7% | +4.7% | +1.7% | 64% |
| Losers bounce back | -3.7% | -3.8% | +5.7% | -1.3% | +0.5% | 48% |
| Winners keep winning | +7.5% | +0.1% | -5.8% | -1.4% | -2.5% | 52% |
| Cheap to bet on a rise | no data that year | -11.3% | -2.9% | -3.8% | +1.8% | 43% |
| Busy coins | -3.9% | +0.4% | -9.0% | -1.7% | -5.7% | 40% |
What this does not mean. The first and last years are partial years, so their averages rest on fewer months.
What should you do with this?
Two kinds of reader use this page. One runs money for other people and has rules to follow. The other is deciding about their own savings. The same evidence leads to different actions.
If you run money for others
Funds, trading teams, the people who manage a company’s cash, research teams.
- You hold as much of the “Few buyers so far” coins as of everything elseThe top and bottom groups are far enough apart to mean something, but the middle groups came out jumbled. So use this to steer clear of the bottom group, not to pile into the top one.
- You want to copy the top group exactlyDo not. On every rule here the middle groups are out of order, which means the exact ranking is noise. Only the two ends are far enough apart to trust.
- You are quoted a trait return without costsAsk how many coins get swapped each month. A trait that swaps most of its coins monthly loses a large share of its advantage to those costs.
If it is your own money
Anyone deciding what to do with their own savings.
- You are choosing which large coins to holdThe traits that paid here are cheap to check yourself: look at what share of everyone with a bet open is betting the coin goes up, and look at whether the pile of open bets on the coin grew over the last month, and look at how little a coin’s price wobbled over the last month.
- You hold coins that sit in bucket 5 of a paying traitThat is the group the trait says to avoid, and on average it trailed bucket 1. It does not mean sell today, but it does mean know why you hold them.
- You want to copy bucket 1 exactlyAround 12 coins, re-sorted every month, with a fee each time, is a lot of work for most people. Use the idea to avoid the worst bucket rather than to chase the best.
The conclusion
A monthly bucket test of 7 simple coin traits across the large, actively traded coins.
Most coin picking is done on stories. This asks the boring question first: which measurable traits have actually paid, and how often.
It tells you which traits deserve a place in your thinking and which sound clever but did not pay, and it shows you how much the cost of buying and selling takes away.
“Few buyers so far” sorted the coins best: its top group beat its bottom group, and it was ahead in 58% of months. Sorting is not the same as making money, though. Over the same period $100 in that top group became $250 while $100 left in Bitcoin became $359. 3 of the 7 traits paid after costs and won more months than they lost. “Busy coins” did not pay.
Words used on this page
Every technical word above is explained again here, in plain English.
- BucketOne fifth of the coins, grouped by how they score on a trait. When the count does not divide by five, the spare coins go to the earlier buckets; bucket 5 is never the one that gets a spare.
- TraitA simple, measurable feature of a coin, like how much it rose last quarter.
- GapWhat group 1 made minus what group 5 made, in the same month.
- TollThe trading cost charged whenever a coin enters or leaves a bucket: 0.10% each way.
- Hit rateThe share of months in which bucket 1 beat bucket 5.
- Equal moneyThe same dollar amount in every coin in a bucket, so no single coin dominates.
- FundingA fee that passes between traders of a coin’s futures (contracts that bet on its price), paid by whichever side is more crowded to the other side.
- FuturesContracts that let traders bet on a coin’s price without owning the coin.
- Turnover or swappedThe share of a bucket’s coins that were swapped out at a month-end.
- Large coinsThe coins with the biggest total market value today (price times number of coins) that trade on a large exchange with at least 400 days of history.
- Betting againstMaking money if a coin falls instead of rises. Traders do this with futures; it carries extra costs that this page does not charge.
- Before costs, after costsBefore costs is the raw gap between the buckets. After costs subtracts the trading toll.
- ReturnHow much money grew or shrank over the month, as a percentage of what was put in.
- NoiseDifferences that are just chance, with no real pattern behind them.
- LongBetting on a rise. A futures account is long when it makes money if the coin goes up.
- EtherThe coin of the Ethereum network, the second-largest coin after Bitcoin.
- ExchangeA marketplace where people buy and sell coins. Prices and trading here come from one of the largest.
- Open betsEvery bet needs someone on the other side, and it stays open until one of them closes it. This is all the bets on a coin that nobody has closed yet, added up. A bigger pile means more people are playing.
- PointsShort for percentage points, and always a subtraction. If group 1 made +3% and group 5 made +1%, group 1 is 2 points ahead. Note that 3% is three times 1%, not 2% more than it, which is why we subtract rather than compare.
Where this page could be wrong
- The coins are the ones that are large today. Coins that collapsed and vanished are not in the history, so every bucket looks better than a real investor would have done, and coins that started trading during the period join once they have enough history. This is also why the test starts in 2022: further back, today’s large coins were the small, beaten-down, wild ones, and any trait that picks those out looks like magic for the wrong reason.
- The “Cheap to bet on a rise”, “Bets piling up” and “Few buyers so far” rules can only be measured for coins people bet on, so each rests on the months and coins where that record exists; the month counts are shown beside every result.
- The bucket 1 minus bucket 5 gap behaves like owning one bucket and betting against the other; betting against coins has extra costs that are not charged here.
- About 4 years of monthly tests is still a short record. Some traits take a decade to prove.
- Prices come from one large exchange. Other exchanges can differ slightly.
- Costs are a flat 0.10% per trade. Real costs vary by coin and by size.
- Everything here is about the past. It describes what paid, not what will.