I started this in July to find out whether a small, rules-based trader can make money on Solana memecoins. Every strategy got written down before it traded, on paper, and the losing ones stayed up. Short version so far: I haven't found an edge. This page sums up what I measured. The full notes are kept private.
Who makes money
Mostly the platforms. Pump.fun took about $981M in gross fees in 2025 and Axiom about $547M. Retail traders mostly don't: the share of wallets in profit was under 50% in 20 of the 21 months from April 2024 to December 2025. Whatever money is left goes to insiders, block-0 snipers and a handful of very fast wallets, and most of them have something an outsider doesn't (early information, or a route outsiders can't use).
What trading actually costs
This was the biggest surprise. My first book assumed a 3% round trip. Measured against live Jupiter quotes, the median was 7.9% on the young pools that book actually picked, and 3.2% across all pools of that age. Pools older than 30 days cost about 0.23%. That one number explains a lot of the losses below.
Most of it is fees, not price impact. I quoted the same pools at three sizes:
| Token | $10 | $100 | $1,000 |
|---|---|---|---|
| PURR | 7.78% | 7.83% | 8.35% |
| LEVERCAT | 8.25% | 8.65% | 10.03% |
| KNOTS | 6.20% | 7.20% | 8.35% |
A $10 trade is far too small to move a $200K pool and it still cost 6.2% to 8.2%. Going up to $100 added 0.05 to 1.0 points, and $1,000 added 0.6 to 2.1 points over the $10 price. The routes I checked went through three venues each, three fees, and young pools charge launch-phase fees that come down as the pool ages. So trading smaller doesn't help. Jito tips aren't in these numbers: the paper books never paid any. Since 15 October each minute-book entry also logs the network's observed tip floor at that moment (the 50th and 75th percentile of landed tips), which shows what tips were going for. It is not a cost the paper trades paid.
Things I tested that didn't work
- My first hourly book, M2. It looked profitable: +16.77R booked over its first 51 trades. Then I re-ran every trade on hourly candles. 11 of 50 had dipped through the stop inside the hour and come back, which the hourly check never saw. A real stop order would have filled. With that and the real cost, the book lands somewhere between 0R and −21R. I didn't put money in.
- Eight tweaks to that book. None made it positive. The biggest effect was about $1.33 per $100 trade, against a gap to break-even of $2.10 to $5.62.
- I looked for better entries in fourteen days of hourly data (31,539 observations). Nothing cleared the 3% cost in both weeks and across more than one token.
- Old, deep pools do go up when SOL goes up. Beyond that the dip rule added +0.3%, t 0.3, which is nothing.
- Copying top wallets: the wallets that ranked best in one month lost 68% per trade out of sample when copied. Their profit doesn't carry over from one month to the next.
- Sniping new launches: no edge at any speed I could model. The best rule still lost about 5% per bet out of sample.
- Buying exchange listings. The move is over within about five minutes of the announcement.
- Daily trend rules on bigger tokens. All 12 lost out of sample. Holding SOL did better.
- Funding carry (long spot, short perp) made 6% to 12% a year out of sample, but the t-stat was 1.0 to 1.7 against the 3.0 I'd set, and about half of it came from one token. Inconclusive.
What's still running
The minute books on the screener page: a flush-and-rebound rule (M5D) and its 12-hour version (M5D2), a rule for the first minutes on the trending list (M5A, no trades yet), plus two books that act as liquidity providers instead of trading (M6E, M6Y). On the LP side the fee is real, about 0.5% a day on busy pools, but the position is half a token, so it mostly moves with the token. Their live results are on the screener page.
Checking a coin yourself
The free checks I run on every coin, roughly in order:
- Find the real pool. Many pools with the same name usually means a relaunch farm.
- Pool older than 7 days, with real locked liquidity. Of the pools under a week old I watched, 71–80% vanished.
- Volume no more than about 10x liquidity. A pool claiming millions in volume on almost no liquidity is wash trading.
- Top holders. If a few wallets were funded by the same source, they're probably one person. The Insider Radar does this check.
- Heavy promo in call channels is a bad sign.
None of this is financial advice. Memecoins mostly go to zero.