Why grid bots die: the boundary problem

Every grid has a lower boundary. Below it you have spent all your allocated capital and you are simply long, holding everything you bought on the way down, with no dry powder and no plan. People handle this three ways and only two of them are honest. Wide boundary, small size per level. You survive further down but each round trip earns less. Slow and safe. Hard stop below the boundary. You accept that the grid has failed and you exit the whole position. Painful and finite. "It'll come back." This is not a plan. It is the default that happens when you didn't choose one of the first two. The uncomfortable arithmetic: the depth your grid can survive and the return it generates per cycle are the same dial pointed in opposite directions. Anyone selling you a grid with great backtested returns and a narrow range is showing you a strategy that hasn't met its bad day yet.
CryptoKarl
30 posts · 0 followers
+ Follow

Related reading

4 replies

GrandpaGrizzly· Jul 2026 ago
"It'll come back" is the most expensive sentence in this business and it has bankrupted better traders than any of us.
D
DrawdownDave· Jul 2026 ago
I have run option three. Can confirm it is not a plan.
H
HalfKelly· Jul 2026 ago
The dial framing is exactly right and it's the same trade-off as position sizing, just expressed differently.
Q
QuietVol· Jul 2026 ago· edited Aug 2026 ago
A grid range is really an implicit forecast of the distribution. You're asserting price stays inside a band, and the probability of touching the boundary grows with how long you leave it running. Most backtests I've seen quietly fix the range using the sample they test on, which is why the breach never shows up in the numbers. If you sized the boundary off something like a rolling volatility estimate instead of a fixed price band, at least the dial moves when the regime does. Do you re-anchor your range periodically, or set it once and let it ride until it fails?
Sign in to reply →
← All automated trading