Hindsight

Your trades already know
what you keep getting wrong.

One trade history. One honest post-mortem. Every finding tied to the exact trades behind it.

Run the post-mortemSee how it works
116 fills41 round trips78% win rate−$1,217 nettech-adjacent −$1,409everything else +$192losers held 2.3× longer2 trades straight after a loss lost $1,497+$1,446 kept by waiting 3h after a lossa 5% stop would have cost $88116 fills41 round trips78% win rate−$1,217 nettech-adjacent −$1,409everything else +$192losers held 2.3× longer2 trades straight after a loss lost $1,497+$1,446 kept by waiting 3h after a lossa 5% stop would have cost $88

measured from the sample history

A post-mortem for traders who never kept a journal

Trading journals only work if you keep one. Hindsight reads the history you already have and tells you the habit that is costing you, once, with receipts.

01

No journal to keep

A CSV, a pasted table or screenshots of your order history. No account, no tagging, no habit to build.

02

Counted, never guessed

Every number is arithmetic on your fills, done before the model is involved.

03

Every claim cited

Each finding names the trades behind it. A trade that isn't in your file is struck.

How it works

Four steps. Only the last one uses a language model, and it only gets numbers it cannot change.

01

Load your fills

A CSV of your tokenized US stock trades. Buys, sells, quantities, prices.

T0001 TSLA buy 2.0161 469.41

T0002 TSLA buy 2.4796 442.06

T0003 PLTR buy 6.3167 183.70

T0004 AMD buy 2.4518 226.89

T0005 AMD sell 2.4518 234.00

T0006 MSTR buy 2.2494 324.46

T0007 MSTR buy 1.9382 310.88

T0008 PLTR buy 7.3284 174.04

T0009 TSLA buy 2.4588 416.30

T0010 MSTR buy 2.0784 297.86

T0011 PLTR sell 13.6450 184.01

T0012 WMT buy 10.4984 109.04

T0001 TSLA buy 2.0161 469.41

T0002 TSLA buy 2.4796 442.06

T0003 PLTR buy 6.3167 183.70

T0004 AMD buy 2.4518 226.89

T0005 AMD sell 2.4518 234.00

T0006 MSTR buy 2.2494 324.46

T0007 MSTR buy 1.9382 310.88

T0008 PLTR buy 7.3284 174.04

T0009 TSLA buy 2.4588 416.30

T0010 MSTR buy 2.0784 297.86

T0011 PLTR sell 13.6450 184.01

T0012 WMT buy 10.4984 109.04

02

Grouped into decisions

Fills become round trips: what you paid, what you added, when you left.

T0014buy368.84
T0016buy352.39
T0017buy336.67
T0021sell297.30
P05 · COIN2 adds down−15.6%
03

Counted in code

Every figure is arithmetic on your data, finished before the model is asked.

78%

win rate

−$1,217

net result

−$1,497

lost on 2 trades opened right after a loss

avg winner+3.9%
avg loser−9.4%
04

Named, and cited

The model names the habit and points at the trades. Invented ones are struck.

Your five worst trades all have two adds below entry.

P06P05P10P99P04P35

P99 is not in the file, so it is removed before render

What it tells you

A finding from the sample history. Every figure computed, every trade cited.

Twice, you opened a position at four times your usual size within two hours of a loss.

P10 came 116 minutes after P04 closed at a loss; P06 came 27 minutes after P05. Both were added to as they fell, and together they lost $1,497, more than your whole net result. Waiting three hours after any loss would have kept $1,445.52.

P04P10P05P06

Run it on your trades

Nothing is stored: no account, no database. The analysis runs once and is gone when you close the tab.

No export in your trading app? Paste the table instead, or take screenshots of your order history. No trades handy at all? The sample is 116 fills from a synthetic trader on real US stock prices, and it runs the full analysis exactly as your own history would.