In October we found the "best" KOL pair to copy. +12.7% a trade. It looked great.
It was 10 trades. We'd picked it out of 4 entry delays times 48 exit rules, so 192 combinations. Take out the 3 best trades and it was −1%.
That's why every result I post says how many trades it's based on. A small number can say anything you want it to.
Our own tests and data. Not financial advice.
Tried it, didn't work: I gave my trading bot faster prices, and it got worse.
My bot reads coin prices from DexScreener, a public chart site most of you probably have open right now. I wanted to know how stale that price was, so I compared 625 of the bot's price checks against the actual trades happening on-chain at the same moment.
The answer was worse than I expected. The typical gap between the chart and the chain was about 10%. Half of all checks were off by more than 10%, and about 12% were off by more than 25%. The chart price was a median 15 seconds old, and 1 in 10 checks were 48 seconds old or worse.
The worst moment: during one rug, the chart showed the coin up 81% while the chain had it down 97%.
So the idea looked obvious. Read the price straight from the chain, exit sooner, lose less.
I replayed the bot's exits with on-chain prices instead. Most of them would have fired earlier.
And it came out worse overall.
The losing trades got out sooner, which helped. But the winning trades got cut early, and that cost more than the losers saved.
My best guess at why: the trailing stop was tuned on the slower, smoother chart price. Raw on-chain price is jumpy, so the same trail trips on noise and sells winners in the middle of a run.
The one version that looked better was using chain prices only for the stop-loss and rug checks. But that was on far too few trades to trust.
So I didn't switch anything. The plan is to log both prices on live positions and decide when there's real data.
What I'd take from this:
Faster isn't automatically better. A change that fixes your worst trades can quietly tax your best ones, and you only see it if you measure both sides.
And if you trade by watching a chart: in a fast memecoin, the price you see can be anywhere from 10 to 48 seconds old. Put your exits on the exchange before you need them. Your reaction time on top of the chart's lag is how a small loss turns into a big one.
Lessons from running our own bot. Not financial advice.
Rough couple weeks in every market.
Still trading daily, keeping losses small.
Memecoins aren't investing. It's closer to gambling, but also a skill game: cheap lottery tickets with way better odds than the real lottery.
Stay exposed to the upside. When things pick up, we ride
KOL or Calls Min 1: on Migrated, that was the *trash* side in our tests. We set KOLs Max 0 and leave Calls blank. Mapping to GMGN's KOL list is unverified, so this is a direction, not a magic name list.
Education, not financial advice.
That is a watch, not a mid-curve buy. Buying those prints failed our profit gate after costs. Wait for a separate, named setup, or don't buy. Education, not financial advice.
Most pumpfun creates never migrate — about 2.8% in our 1.56 million-create sample (1,559,895 regular SOL-quoted creates; 2.82% migrated ever). Almost all of the rest are simply dead by hour one, even if they haven't "failed" a filter.
Among coins still *alive* at 1 hour (still printing trades), a simple cut — curve visibly ≥ about 70% — later migrated 39 / 46% in the two halves (38.6% / 45.5%), vs a 6.5% base at that checkpoint.
@genwealthXD If everyone does this, the people who pay the highest fees/tips win. My buy slippage is 3%, priority fee and validator tip both set at 0.00001SOL, MEV OFF (the low slippage makes sandwich attacks impossible).
Even stacking a dev-sold skip and a trailing stop only brought the >300% bucket to about −8 to −11%, still a loss. The chart looks like you're late to a runner. On this tape, you were late to a dump. Education, not financial advice.
Hand-buying a 100–300% jump in the first 5 minutes, when the coin is 1–3 minutes old, lost about 23–26% a trade in our tests — worse than names with no jump (about −9 to −10%). Jumps over 300% also lost, just a bit less.
Check the full CA, character for character, against the project's own official post. Name, picture, and ticker prove nothing. Never copy a CA out of a reply.
Education, not financial advice.
When a $100k name is running, the copies arrive in minutes. Same ticker is not the same coin.
We measured that exact kind of clone as a *buy*: about −33% to −40%. Being first didn't mark the winner either — in one "roadster" cluster of 95 mints, the one that ran was #91.
@crbpite8 It's a game. Memecoins have nothing to do with "investing" or "narrative", the meme market is just a PVP game with no rules played for real money.