ABCDEFGHIJKLMNOPQRSTUVWXYZ
1
Backtest Bias Audit Worksheet
2
Education only; not investment advice. Every example number and timestamp is INVENTED. This worksheet checks the order of events and the arithmetic you type in. It cannot tell you whether your data, rules or results are correct.
3
4
Sheets
5
Same-Close Example: the article's invented rule. It compares an invalid same-close fill with a valid next-open fill for one trade.
6
Trade Timing Audit: one row per trade (up to 30). Enter when the information became available, when the signal was generated and when the fill happened. It flags trades whose information was not yet available at the signal, or whose fill was not after the signal, and shows P&L with and without those trades.
7
Cost Sensitivity: gross P&L, trade count and cost per round trip give net P&L at several cost levels and the break-even cost.
8
Data Split: splits N observations in time order into a research (training) sample, a validation sample and a locked test sample.
9
Audit Checklist: ten gates. A 'Yes' needs evidence written in the evidence column. Any 'No' means reject, or revise and retest on a new locked test sample.
10
Tests: the checks this draft was tested against.
11
12
Conventions
13
Timing rule: information time ≤ signal time < fill time. A fill at the same moment as the signal is flagged, because a signal that needs a candle's close cannot be filled at that close.
14
Enter times as real date-times (for example 2026-03-02 15:30). Do not type text. Use one time zone and say which in your notes.
15
P&L per trade = (exit − entry) × units for a long, (entry − exit) × units for a short. Costs are NOT included in the Trade Timing Audit; use Cost Sensitivity for costs.
16
Net P&L = gross P&L − (number of round trips × cost per round trip). Break-even cost per round trip = gross P&L ÷ number of round trips.
17
18
Limits
19
'Bias-free' is a goal, not a certificate. Passing every gate reduces known flaws but cannot remove data errors, model uncertainty or future regime changes, and it does not predict live results. Paper fills remain simulated.
20
Draft not yet opened by a person in Excel or Google Sheets. Check the formulas before relying on them.
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100