Community › Strategies & Backtesting

my rule of never sizing off the entry, only off the invalidation

CryptoKarl· May 2026 ago· 👍 8· 💬 8
ok so this is the one thing that actually dragged me out of the blow-up-every-3-months loop and i wanna hear if anyone does it differently. for years i sized like a degen: "i want 5x on this" and then i'd pick a stop wherever it felt vaguely safe. which is backwards. the leverage number was the input and the risk was whatever fell out the other end. some trades i was risking 0.4% of the book, some i was risking 6% and had no idea until it was already red. now i do it the other way. before anything i mark the level that makes the idea wrong. not "where it hurts" — where the reason i entered stops existing. then i decide the flat dollar amount i'm ok setting on fire, which for me is a fixed % of the account and it does not move because i'm feeling confident. size = risk / distance to invalidation. leverage is just whatever number the exchange shows me after that, it's an output, i genuinely don't look at it anymore except to check i'm not near liq before the stop. side effect i didn't expect: it killed a whole category of bad trades. if invalidation is really far away, the position comes out tiny, and if the position is tiny enough that a win wouldn't matter, i just... don't take it. the math tells me the setup is bad before my ego does. stuff i'm still bad at: - funding. i hold perps for days sometimes and i still treat funding as a rounding error when on longer holds it clearly isn't. do you bake carry into the risk number or keep it separate? - scaling in. every time i add to a winner i move my average and my original invalidation stops being the real invalidation, and then i'm improvising. no clean system for this yet. - weekends. thin books, my stop distance assumptions are basically fiction, so lately i just carry less into them. anyway curious how others handle sizing. anyone size off ATR instead of a structural level? i tried it and it felt too mechanical for crypto but maybe i did it wrong.
CryptoKarl
30 posts · 0 followers
+ Follow

Related reading

Strategies & Backtestingmy two-strike rule and why i keep breaking it anywayBrokers & APIsmy kill switch does more for me than any entry tweakBrokers & APIsmy kill switch rules for perps, curious how others do itStrategies & Backtestingthat dip-fading writeup — does the asymmetry even survive on perps?

8 replies

N
NightOwl_Yuki· May 2026 ago· edited Aug 2026 ago
the invalidation-first thing is the same conclusion i came to but it took me way longer lol. on funding. I keep it separate but i do check the rate before entering anything i think i'll hold past a couple of fundings, and if it's ugly on my side i just want a better entry to compensate. for the scaling in problem, what helped me was deciding upfront that an add is its own trade with its own invalidation, not an upgrade to the first one. keeps me honest even if the average price looks weird.
L
LurkerLee· Jun 2026 ago
when you say the position comes out too tiny to bother — do you have an actual cutoff number for that, or is it a feel thing?
DataDrivenDee· Jul 2026 ago
the part i'd want to poke at is the "i just don't take it" filter. do you log the ones you skip? without that you'll never know if the tiny-size trades were actually bad or just wide, and those are different problems. i've caught myself skipping perfectly good slow setups because the size looked unexciting.
CryptoKarl· Jul 2026 ago
@LurkerLee it started as feel and that was useless, so now it's roughly: if a full move to my target moves the account less than a bad day of funding, it's not a trade it's a hobby. and yeah @DataDrivenDee i do not log skips, which you have now made me feel bad about, thanks. adding it.
R
RiskFirstRita· Jul 2026 ago
The scaling-in problem isn't a sizing problem, it's a bookkeeping problem. When you add, you have two positions with two different invalidations, so treat them as two. Original tranche keeps its original stop, new tranche gets its own. If you can't state both without doing mental arithmetic, you've added too fast. And please keep doing the weekend thing.
H
HalfKelly· Jul 2026 ago· edited Aug 2026 ago
On ATR vs structure, they're answering different questions. ATR gives you an average distance the thing wanders, structure gives you the distance at which your reason dies. If your reason is structural, size off structure and use ATR only as a sanity check: if your invalidation is inside one ATR, you're going to get stopped by noise regardless of how right you are. On funding, the clean way is to treat carry as a reduction in expected value, not as risk. Risk is the stop. Carry just shrinks the payoff, so a trade that only works at zero funding was never much of a trade.
BacktestBetty· Jul 2026 ago· edited Aug 2026 ago
Couple of things from testing this kind of rule: 1) fixed fractional risk off invalidation is very hard to overfit, which is exactly why it survives, there's basically one parameter. 2) when I tested ATR-based stops vs structure stops the ATR versions looked better in backtest and worse live, and I'm fairly sure it's because ATR stops are trivially easy to tune the multiplier on until the equity curve looks nice. 3) weekends: I'd actually test that assumption rather than assume it, split your results by session and see if your slippage is really worse or if it just feels worse.
M
MomoQueen· Jul 2026 ago· edited Aug 2026 ago
the add-to-winner thing clicked for me when i stopped calling it an add — if the new tranche can't stand alone as a trade i'd take fresh, with its own reason and its own invalidation, it's not an add it's me being greedy! kills like 80% of my scale-ins instantly. also on weekends, same as you, i just carry smaller, but honestly the bigger fix was accepting my stop is a request not a guarantee when the book is thin. I size like the fill could be worse than the level and if that makes it too small to bother, cool, that's information!
Sign in to reply →
← All strategies & backtesting