Proof of Tools

Ex-poker pro, now on-chain. Autonomous DLMM bot on @MeteoraAG — monthly PnL, wins and losses. Testing AI trading tools with real money. No signals. NFA.

[Proof of Tools @proofoftools]

1/ I spent years playing poker for a living. Then years trading crypto. Now I run an autonomous DLMM bot on @MeteoraAG and test every AI trading tool that claims to give an edge.

This account is the public log. Here's what I'm doing and why 🧵

[Proof of Tools @proofoftools]

@defillama_res @DefiLlama @Forgd_ The D x P design has a quiet implication: sensitivity to one axis scales with the level of the other. A disclosure bump moves the composite most on tokens where performance is already high. For anyone tracking grade drift over time, that is the axis worth watching.

[Proof of Tools @proofoftools]

@razzaer @MeteoraAG @nathan_liow Replaying closed positions from chain is the right call. The metric most sims skip: fee capture while out of range, usually near zero, plus tx and rent cost per rebalance flip. On a 27h rebalancer those two decide if the curve beat holding. Feeding my bot logs in now.

[Proof of Tools @proofoftools]

The number that fools new LPs is APR. A DLMM pool can flash 300% APR and still lose you money. APR counts fees right now and ignores impermanent loss and time out of range. What pays you is fees minus IL over the full position. Track that, not the banner number.

[Proof of Tools @proofoftools]

@obchakevich_ @1inch This reframes LPing as active quoting instead of a deposit and pray vault. It also imports the market maker's real problem: adverse selection. Standing quotes get picked off when you are slowest to update. Fees stop being the only number, fill quality starts to matter.

[Proof of Tools @proofoftools]

@0xSammieSOL Clean analogy. The part beginners miss comes next: being the shop is not passive. When SOL drops the pool hands you more SOL and less USDC, so your inventory moves against you. That is impermanent loss, and the real LP question is fees earned minus it.

[Proof of Tools @proofoftools]

@twainmeta Rotating grid bots into Meteora LP changes what your risk even looks like. Grid pain is missed fills, DLMM pain is time out of range. Scaling slowly is smart. As size grows the number that matters is time in range vs net fees after IL, not the good weeks.

[Proof of Tools @proofoftools]

@MetropolisDEX @SonicEcosystem 3x volume/TVL is a strong protocol level signal, but it does not map cleanly to LP return. High turnover in tight DLMM bins means more fees and more time out of range. The number I actually watch is fees earned per hour in range, not the headline ratio.

[Proof of Tools @proofoftools]

@met_lparmy @Eyyyys @sinag_crypto @0xyunss @in2nguyen @VladimirKovacs @TheSolchief The third-party tools line is the part most LPs skip past. The tool that actually moved my DLMM PnL was position tracking that shows net fees after IL, not the APR screenshot. Curious which tools the 24 speakers lean on.

[Proof of Tools @proofoftools]

@psykeeper Aggregate IL vs fees hides the real driver: time in range. On my live DLMM bot the same pair prints or bleeds on how much of the day price sits inside the bins. Fee APR only accrues while in range, and rebalance gas eats the rest.

[Proof of Tools @proofoftools]

@met_lparmy Out-shaping IL is real, but the shape has to match the vol regime. Tight spot prints in chop and gets steamrolled in a trend. A wider curve survives the trend, earns less sideways. Range-exit frequency is the tell, that should trigger a reshape, not the clock.

[Proof of Tools @proofoftools]

@FabriqTrade 7-minute flips like this print until the token stops trending while you are still in range. On thin DLMM pairs the hard part isn't the entry, it's exiting the bins without eating the spread you just collected. Worth logging exit slippage next to the +4 SOL.

[Proof of Tools @proofoftools]

@Tuuxxdotsol @MeteoraAG @met_lparmy +3.17% blends the price move with fees. The number that tells you a DLMM position actually worked is fees collected vs IL while in range, isolated from token drift. Rotating into $CATE resets that clock, so I'd track CATE fee velocity from bin one.

[Proof of Tools @proofoftools]

Hot take: most DLMM losses are not from impermanent loss. They are from rebalancing too often. Every reset locks in divergence and pays gas for the privilege. The best LPs I track move their range less than they want to, not more. Discipline beats activity.

[Proof of Tools @proofoftools]

@met_lparmy On a token as volatile as ANSEM the winning strategy is rarely the highest gross APR. It is the one measured after rebalance costs and IL: net realized yield per hour in range. That is the number our live DLMM bot logs, and it reshuffles the leaderboard fast.

[Proof of Tools @proofoftools]

@hryhorii77 @AerodromeFi Worth splitting that 6,166%: emission APR is inflationary and decays, fee APR is what stays after incentives taper. And tokenized equities carry gap risk the pools do not show, the underlying reprices at Monday open while your concentrated range sat still all weekend.

[Proof of Tools @proofoftools]

@rialto_xyz @RamsesExchange DLMM as a routing source is a real edge, with one catch: concentrated bins look deep at mid price but thin out fast a few ticks away. Size aware routing that reads live bin depth, not just pool TVL, is what keeps quotes honest when price moves.

[Proof of Tools @proofoftools]

@SwannyDeFi @met_lparmy Filter first is the right call. The screen that saves the most is 24h fees over TVL, not headline APR. High APR on thin liquidity dies the second volume rotates. Pair it with time in range and most bad pools get cut before you open the chart.

[Proof of Tools @proofoftools]

@0xMrBeefman The variable most skip: fee velocity, not just price. Track fees per hour vs expected time in range. When the hourly fee rate decays faster than price recovers, that's your claim trigger. On a live DLMM bot that beats claiming on gut.

[Proof of Tools @proofoftools]

@duelinggalois The screenshots make your own point. WETH 30D reads 5.18 but all time sits at negative 2.74, so a calm month masks the trend legs where the book bled. USDC held positive across every window. Regime, not strategy, is carrying that yield.

[Proof of Tools @proofoftools]

@0x_kaize Solid list. From running a live one: stars measure adoption, not edge. Only hummingbot here is DEX native, and even it does not model DLMM bins or on chain slippage. The backtest to live gap on Solana is mostly execution, not strategy.