Report the distribution, not just the summary

The summary gives me totals and averages. Averages are the least informative statistic for a trade distribution with any skew, which is most of them. What I'd like alongside: median trade, the shape of the distribution, and the result with the top and bottom few trades removed. That last one is a check several people here recommend running manually. Building it in would make it something everyone does rather than something the careful ones do. Not more numbers for their own sake. These three specifically, because they answer "is this real" better than the total does.
Q
QuietVol
13 posts · 0 followers
+ Follow

Related reading

6 replies

DataDrivenDee· Jul 2026 ago
Building the leave-out check into the default view would probably improve this community's research quality more than any other single change.
BacktestBetty· Jul 2026 ago
Strong agreement. The checks people skip are the ones that require an extra step.
TraderWeTraderWe Team· Aug 2026 ago
Noted, and it's a fair ask. Median and a trimmed result are straightforward to compute from data we already hold. The distribution shape is the harder one — a histogram needs bucket choices and screen space, so if we do it, it'll likely start as something simple like quartiles rather than a chart. No timeline promise, but it's on the list.
F
FiveMinFiona· Aug 2026 ago
One thing I'd add: I'd want the trimmed number labelled clearly with how many trades were dropped. Removing the top and bottom 3 from a sample of 20 is a very different statement than from a sample of 400, and people will screenshot the number without the context.
I
IndicatorSkeptic· Aug 2026 ago
Funny how nobody ever asks for the trimmed number when they're up. It only becomes interesting statistics when it flatters you.
Q
QuietVol· Aug 2026 ago· edited Aug 2026 ago
IndicatorSkeptic. That asymmetry is exactly why I want it in the default view rather than as an option. An optional check gets run selectively, and selective checking is worse than no checking because it produces a biased impression of rigour. FiveMinFiona's point on labelling the drop count is well taken; a trimmed mean without the sample size is nearly meaningless. Would people find quartiles sufficient, or is the full shape actually what you're after?
Sign in to reply →
← All feature requests