i stopped trusting single-exchange candles and it fixed a lot

ok so this is more of a data hygiene rant than a strategy post but bear with me. for like a year my whole perp process was built on candles from one venue. entries, my ATR stop sizing, everything. and every so often i'd get these trades where the wick on my chart clearly tagged my level and bounced, but the fill was garbage or the move just... didn't exist anywhere else. one wick, one venue, gone in a second. i'd size up on what i thought was a real sweep and it was basically one guy nuking a thin book at 4am. what i do now, and honestly it's dumb simple: before i take anything based on a wick or a sweep, i eyeball the same window on two other venues. if the wick is only on one chart it isn't a level, it's noise, and i either skip it or i cut my size in half. i also stopped putting hard stops at round numbers that only show up on my primary feed for the same reason. the other half of it is funding and OI, which i treat as its own data problem. i pull those separately instead of just squinting at price, because price alone told me nothing about how crowded a move was. i don't trade off funding directly, it's more like a size dial - if everyone's paying to be long and i want to be long too, i'm going in smaller than i'd like to. things i still haven't figured out and would take input on: - how much do you care about candle timestamp alignment across venues? like some feeds close the minute differently and i've definitely fooled myself comparing two charts that weren't actually the same minute - do you look at aggregated/index price at all or is that just extra abstraction - volume across venues is basically uncomparable to me, wash trading and different fee structures, so i mostly ignore volume numbers and just use them for relative shape on a single venue. is that too paranoid the degen part of my brain hates that this process makes me pass on stuff. but the trades i passed on were mostly the ones i'd have been stopped out of by data i shouldn't have trusted.
CryptoKarl
30 posts · 0 followers
+ Follow

Related reading

4 replies

N
NightOwl_Yuki· May 2026 ago
the single-venue wick thing is very real on the night session, thin books do weird stuff and it looks so convincing on the chart until you check anywhere else. i do basically the same two-venue check, mostly out of laziness i just have them stacked in one window.
MLcurious· May 2026 ago
the timestamp thing is exactly what wrecked my first attempt at building a dataset lol. i merged candles from two sources on the timestamp column and didn't realize one was labeling the bar by open time and the other by close time, so everything was shifted by a bar and my "model" was just quietly peeking at the future. took me ages to notice because the results looked great. now i check what the timestamp actually means before anything else. also curious about the volume question, is relative shape on one venue really enough or do people normalize it somehow?
F
FiveMinFiona· Jul 2026 ago
on the timestamp question - it matters more than people think and it's cheap to fix. i label every bar by open time, write it down somewhere, and check any new feed against a known bar before i use it. on your third point i think you're fine ignoring cross-venue volume levels, but don't throw away shape on your primary - relative volume within one venue is still the most useful part of the candle for me.
Q
QuietVol· Jul 2026 ago
index price is worth a look precisely because it's a distribution summary rather than a single draw - a wick that survives the aggregation had size behind it somewhere, one that vanishes was a tail event on one book. i wouldn't trade off it, more as a filter on whether the level exists at all. on funding as a size dial, do you scale it continuously or just have a threshold where you halve?
Sign in to reply →
← All market data & research