Six ideas, tested.

Question

When Binance, Bybit, OKX, Coinbase or Upbit announces a new listing, can you buy the coin where it already trades and sell into the listing?

Data

1,043 listing announcements from April 2025 to October 2026, with one-minute prices from wherever each coin already traded. Fees of 0.2% a side on exchanges and 1% on DEXs, plus slippage from 0.05% to 1% a side depending on the coin's volume.

Test

The oldest 730 announcements to build, the newest 313 (from 16 April 2026) to judge. Thousands of entry, exit and filter combinations searched on the first set only.

What we found

Speed is the whole edge. A minute after the post, small coins with little trading still had room to run. By the time most people have read the post, the average trade loses money.

The best strategy's likely range in the test period was wide: +1.6% to +20.9% a trade, at 90% confidence. About 29% of its trades won, and the winners did the work.

Searching thousands of combinations always turns up some that look great by luck. Of the 2,670 that made money in the build period with enough trades, 42% lost money in the test period. The fairest summary of the whole search: the 10 strategies we picked on the build period alone averaged +1.8% a trade in the test period, and 5 of them lost money. The strategy in the table was the best of those 10.

Weak spots

• 48 trades is a small sample.
• Getting in within a minute takes automation and an account on the venue where the coin already trades.
• Binance and Bybit block US connections, so we used their public data mirrors and other venues' prices.

Question

New coins launch with hype and a lot of unlocked supply waiting to be sold. Do they trail bitcoin afterwards, and can you short that?

Data

Every Binance listing from January 2020 to September 2026 with 90 days of prices, delisted coins included. For the short test, only coins that had a Binance perpetual future by the entry day, with the funding they actually paid or received. Prices are Binance spot.

Test

Listings up to the end of 2022 to build, 2023 onward to judge.

What we found

The drift is real and got stronger after 2022. Shorting every new Binance perpetual the day after listing, holding 90 days, hedged with bitcoin, 3% of the account per short, with a stop if a short went 100% against us: 189 trades from 2023 on, +26% a year, worst drawdown 27%.

Then the costs. Until 2024, shorts on these coins usually got paid funding to hold. A 90-day short opened in 2025 paid about 9% in funding on average. In 2026 it was about 22%. That's what a crowded trade looks like.

Without a stop, single shorts lost as much as 684%. The stop isn't optional.

Weak spots

• Only 23 shorts in the build period, because Binance listed few new perpetuals before 2023.
• US residents can't use the perpetual futures venues this needs.
• Rising funding could erase the edge going forward.

Question

Does the classic trend-following rule work on altcoins: buy a coin when it closes above its highest price of the last 20, 50 or 100 days, and ride it with a trailing stop?

Data

677 Binance USDT pairs since 2017, including 194 that no longer trade. 288 combinations of lookback, volume filter, bitcoin trend filter, stops and position count. Fees of 0.3% a side plus slippage by volume.

Test

Tuned on 2018 to 2022, judged on 2023 to October 2026. The ten best settings were picked on the first period before we looked at the second.

What we found

The training winners made their money in the 2020 to 2021 altcoin run: BNB +640%, CHZ +505%, EGLD +399%. From 2023 on, 19% of the 288 settings made any money and none beat just holding bitcoin, on return or on risk.

At double the fees the best setting did slightly worse (−31% a year). Limiting it to coins Coinbase or Kraken list today helped a little (−18%) and still lost.

Weak spots

• Trend following earns its money in big trending runs. Another altcoin season could make it look good again. We can't test one that hasn't happened.

Question

Some pump.fun tokens pump, dump 40 to 60%, then pump again. Can you catch the second leg without racing the sniper bots for the first?

Data

93,432 setups from 19,082 tokens in their first hour after launch, from our own archive of every pump.fun trade between late February and mid March 2026. Entry 30 seconds after the bounce. Costs of 5% a side.

Test

The first 70% of days to build, the last 30% to judge. 5,856 entry and exit rules tried.

What we found

Second legs happen often enough to notice on a chart and nowhere near often enough to pay for trading them. Even before costs, the best rules earned under 2% a trade.

One signal held in both periods: tokens that bounce within about a minute of their peak double more often (23% vs 13% in the test period). It still loses after costs.

A data trap worth knowing: about 4,300 tokens traded below the standard pump.fun price curve (a different pool or a non-standard curve). Left in, they made one rule look like +67% a trade. We removed them.

Weak spots

• The archive follows each token for its first hour only.
• Real costs depend on size and slippage. 5% a side is our estimate.

Question

Exchanges often move a token into their own wallets before they announce it. Can you see that on-chain and buy first?

Data

Transfers into publicly labelled exchange wallets on Ethereum and Base, matched against the listing announcements from study 01. Bought when the wallet first touched the token, sold 5 minutes after the post. About 1.2% round-trip cost.

What we found

The tell is real for some exchanges. Binance and Upbit move tokens after they announce, so there's nothing to see. Where a tell does exist, the listing barely moves the price, so there's little to catch.

Weak spots

• OKX and Bybit have too few trades to count by our own rule. Bybit's small positive number is noise until proven otherwise.
• Wallet labels are public and incomplete, so some tells were missed.

Question

If trading bots take profit at round multiples like 2x or 3x, prices should stall there, and you could sell just ahead of them. Do they?

Data

5.1 million completed buy-then-sell positions on pump.fun, from the same archive as study 04.

What we found

Exits don't bunch up at 1.5x, 2x, 3x or 0.5x. Whatever bots are doing, it isn't visible as round-number selling.

What we did find: 7.2% of round trips are volume bots. They buy and sell at the same price about 13 seconds apart, most likely to make a token look busier than it is.