Been grinding on a Solana trading bot for about 6 months now, getting pretty close to the finish line. I’m mainly here to see if anyone else is working in the same space, exchange ideas, maybe network with people who are deep into this.
Couple questions first:
I’m looking for people who are either very experienced developers or just full-blown crypto.
Quick summary of what I’ve built so far:
Real-time execution, consistently under 200ms, streaming straight from the chain. Every token gets scored with a mix of velocity, pressure, momentum, liquidity changes, and a few normalized activity ratios. The scoring system uses distribution logic so only the strongest signals scale into bigger trades, weaker ones get rejected.
Sizing isn’t fixed — it flexes up and down based on how well the trade lines up with what’s been working recently. Think small for weak signals, bigger for strong pattern clusters, and max size only when conditions are almost identical to past winners.
There’s an adaptive layer that tracks recent trades, figures out which combinations actually hit, and boosts those patterns while decaying stuff that stops working. It’s fully self-adjusting, no hardcoded rules.
All trades are built manually from raw chain data, no SDK shortcuts, so I’ve got low latency and access to coins before they show up anywhere else.
Also added a defense system — after a string of bad trades it automatically tightens everything, cuts size, and waits until recovery is proven before scaling back up.
Right now I’m finishing the “auto switching filters” and exploratory mode (basically tests off-pattern trades to discover new winning clusters). Not fully done yet.
Best session so far was around 73% win rate, average win about 4x bigger than average loss.
Just curious who else is in this space. Anyone running similar setups, or even crazier ones? Happy to talk shop with people at the same level.
Couple questions first:
- Anyone here using Helius Laserstream instead of the usual gRPC node? I’ve been testing with gRPC and curious if Laserstream really gives any noticeable edge in timing.
- Has anyone managed to squeeze better reaction times than ~100–200ms end-to-end?
- Anyone here actually building raw transactions on-chain manually, pulling in the right accounts/keys instead of relying on SDKs?
I’m looking for people who are either very experienced developers or just full-blown crypto.
Quick summary of what I’ve built so far:
Real-time execution, consistently under 200ms, streaming straight from the chain. Every token gets scored with a mix of velocity, pressure, momentum, liquidity changes, and a few normalized activity ratios. The scoring system uses distribution logic so only the strongest signals scale into bigger trades, weaker ones get rejected.
Sizing isn’t fixed — it flexes up and down based on how well the trade lines up with what’s been working recently. Think small for weak signals, bigger for strong pattern clusters, and max size only when conditions are almost identical to past winners.
There’s an adaptive layer that tracks recent trades, figures out which combinations actually hit, and boosts those patterns while decaying stuff that stops working. It’s fully self-adjusting, no hardcoded rules.
All trades are built manually from raw chain data, no SDK shortcuts, so I’ve got low latency and access to coins before they show up anywhere else.
Also added a defense system — after a string of bad trades it automatically tightens everything, cuts size, and waits until recovery is proven before scaling back up.
Right now I’m finishing the “auto switching filters” and exploratory mode (basically tests off-pattern trades to discover new winning clusters). Not fully done yet.
Best session so far was around 73% win rate, average win about 4x bigger than average loss.
Just curious who else is in this space. Anyone running similar setups, or even crazier ones? Happy to talk shop with people at the same level.
