What I check after any broker or platform update

Updates are the most common cause of a setup that worked yesterday and doesn't today. Four things, in order: 1) Reconnect and confirm a test fill, not just a status light. 2) Check that my order types still behave the way I assumed. 3) Confirm the data I depend on is still arriving at the same rate. 4) Run one session small before running it normal. Item 4 is the one that requires patience and it's the one that has saved me. An update that changes something subtle won't announce itself; it'll just show up in your fills a week later as a mystery.
P
PremarketPete
20 posts · 0 followers
+ Follow

Related reading

10 replies

BacktestBetty· Jun 2026 ago
"Run one session small" is underrated. It's the cheapest possible A/B test against your own assumptions.
GrandpaGrizzly· Jun 2026 ago
Learned this the hard way more than once. The subtle changes are the expensive ones.
N
NightOwl_Yuki· Jun 2026 ago
crypto side has the extra fun of updates landing mid-session because there's no close. i keep a note of the last version i traded normal size on, so when things feel weird i at least know if something changed under me.
H
HalfKelly· Jun 2026 ago
Item 4 is basically sample size with a smaller unit. One session small tells you almost nothing statistically, but it's cheap and it catches the loud failures. The quiet ones you only find by keeping a log of fill quality over a few weeks and comparing before/after.
SlowSwing_Sam· Jul 2026 ago
One of the perks of holding for days is I usually find out about the broken thing while I'm making coffee instead of mid-scalp. Still check the order types though, that one bit me once.
MLcurious· Jul 2026 ago· edited Aug 2026 ago
Dumb question maybe. How do you actually confirm data is arriving at the same rate? Is that just eyeballing timestamps or do you log it somewhere? I wouldn't even know what normal looks like for mine.
R
RiskFirstRita· Jul 2026 ago
Add a step zero: write down what you expect each thing to do BEFORE you test it. Otherwise you look at the result and decide that was what you expected all along. And nobody should be running full size the first session back, that's not caution, that's just the baseline.
M
MomoQueen· Jul 2026 ago
The mystery fills line is so real!! Nothing worse than losing a week to slippage you can't explain and then finding out it was a setting. I keep size dialled down for a day after any update no matter how good the setup looks.
DataDrivenDee· Jul 2026 ago· edited Aug 2026 ago
To MLcurious's question. I log message timestamps to a file and just look at the gaps. You don't need to know what "normal" is in the abstract, you only need your own baseline from before the update. Takes one boring week of collecting to have something to compare against. Pete, do you actually measure rate or is it more of a feel thing?
S
SpreadWatcher· Jul 2026 ago
Order types are the one I'd put first, not second. A limit that silently becomes marketable, or a stop that starts routing differently, shows up as spread paid rather than as an error message. Nothing breaks, you just bleed.
Sign in to reply →
← All brokers & apis