The data you have shapes the strategies you can imagine

An underrated constraint. If all you've ever seen is daily bars, you will invent daily-bar strategies, and the space of ideas available to you is quietly bounded by your inputs. This cuts both ways. Richer data opens ideas. And it also opens a great many ways to fool yourself that coarser data simply doesn't offer. I have watched people go from a working simple system to an elaborate broken one purely because more data became available to them. Get the better data. Then be more suspicious, not less.
GrandpaGrizzly
16 posts · 0 followers
+ Follow

Related reading

8 replies

BacktestBetty· Jun 2026 ago
"More data opens more ways to fool yourself" belongs in every research checklist.
I
IndicatorSkeptic· Jun 2026 ago
Also explains why the fanciest setups so often underperform the simplest ones.
H
HalfKelly· Jun 2026 ago· edited Aug 2026 ago
The degrees-of-freedom point is really a sizing point too. Every extra parameter you fit shrinks your effective sample, and your edge estimate gets noisier. Which means the Kelly fraction you *think* you computed is biased high. Same reason people use half Kelly: you're not sizing your edge, you're sizing your uncertainty about it.
D
DrawdownDave· Jun 2026 ago
I went from daily to minute bars once and immediately invented a strategy that needed the 3rd minute of the hour. It worked beautifully until it didn't, and I never could explain why it worked in the first place, which should have been the tell on day one.
R
RiskFirstRita· Jul 2026 ago
Add it to the checklist: any new data source gets a written note of what you expect it to improve *before* you look. If you can't say why intraday helps your idea, you're just shopping for confirmation. And please don't upsize because the new backtest looks prettier.
DataDrivenDee· Jul 2026 ago· edited Aug 2026 ago
Genuinely curious how you'd test the claim, GrandpaGrizzly. Has anyone tried building the same idea on daily and then on finer bars and compared out-of-sample decay? Seems measurable in principle but I'm not sure what you'd hold constant.
S
SpreadWatcher· Jul 2026 ago
Finer data also shows you all the little wiggles you can't actually trade. Half the ideas that appear at the minute level are just spread bouncing back and forth, and the fill assumptions do the rest of the lying.
MLcurious· Jul 2026 ago
This hits home, I got tick data and immediately wanted to throw a model at it. Honest question though — is the fooling-yourself risk about the data itself or about the model complexity it tempts you into? Still fuzzy on where the line is.
Sign in to reply →
← All market data & research