RealTest · daily signals matched to the backtest

Signals & positions

RealTest · the signals of the 3 years before the forward test, replayed in the backtest

Backtest signals

RealTest · results read live from your Scripts folder

Lumi

Loading results…
Costs Variant picked Scored Pass
Strategy

Period
or drag across the chart · double-click to reset

Calendar-year returns · latest year to date

Base version and two optimisation attempts

Did the selection period predict the test period?

Each dot is one strategy's chosen variant. Dots above the green line passed; dots on the dashed diagonal held their Sharpe. Click a dot to open it.

Test-period Sharpe across all strategies

Overfitting check · CSCV

The variants’ shared history is cut into equal blocks of trading days. For every way of splitting the blocks half and half, the variant with the best Sharpe in one half is the pick, and it is ranked against all the variants in the other half. PBO is how often the pick lands in the bottom half.

Point-in-time universesStock strategies use Norgate's "Current & Past" index lists, delisted stocks included, and a stock only counts on days it was in the index. The liquidity-ranked re-tests (#401–420 and #471–480, "ever-member pool") rank stocks that were in an index at any time, so a few future joiners slip in: read those results as slightly optimistic. The earlier results on today's index members are kept in Scripts\archive and shown dashed as "Old data" (use Hide old at the top to hide them).
Selection luckEvery strategy tried makes a lucky pass more likely. Judge a pass together with how many ideas were tested: the overfitting check above measures it for whatever is in the list.
CostsEach script sets its own slippage and commission. The research batches use 10bp per fill, which is conservative for liquid large caps.