Models

GARP Core 10

An open-source trading model based on GARP principles that runs in Claude or ChatGPT.

Run in Claude Run in ChatGPT View on GitHub

GARP Core 10 returned +573% over the backtest, outperforming the S&P 500's +149% and ending 2.7× its result.

Stats

model
6.73×
S&P 500
2.49×
statS&P 500model
$24,907 $67,310
+149% +573%
+14% +32%
-34% -18%

GARP Core 10 grew faster than the S&P 500 (+32% vs +14% a year), with a shallower worst fall (−18% vs −34%).

Edge

model beats S&P 500 by
2.70×
statedge
2.82×
0.96×
98% of days
+16%

GARP Core 10 beat the S&P 500 by 2.7× and was ahead 98% of the time.

Yearly ROI

'20
'21
'22
'23
'24
'25
'26
yearS&P 500model
2026 +8.7% +11%
2025 +18% +45%
2024 +25% +28%
2023 +26% +21%
2022 -18% +28%
2021 +29% +30%
2020 +17% +62%

GARP Core 10 beat the S&P 500 in 6 of 7 years; its best year was 2020 at +62%.

Rolling starts

yearly returnS&P 500model
+16%/yr +28%/yr
+22%/yr +34%/yr
+11%/yr +23%/yr
100%

Whichever week GARP Core 10 started, its yearly return to date ran between +23% and +34%, above the S&P 500's from every one of 253 starts.

Holding periods

1 year from each start · 306 starts · beat S&P 500 in 77%

2 years from each start · 253 starts · beat S&P 500 in 91%

3 years from each start · 201 starts · beat S&P 500 in 100%

held for S&P 500 median model median
+18% +31%
+37% +69%
+45% +116%

When started on any week, GARP Core 10 beat the S&P 500 100% of the time, at a median +28% a year against +16%.

Trades

ticker entryexit roi
ALL Aug 21 Sep 11 +0.3%
APP Aug 21 Sep 11 +3.6%
B Aug 21 Sep 11 -6.9%
CDE Aug 21 Sep 11 -4.4%
FAF Aug 21 Sep 11 -0.3%
GME Aug 21 Sep 11 +16%
INDV Aug 21 Sep 11 -8.6%
MCY Aug 21 Sep 11 -0.5%
PAAS Aug 21 Sep 11 -3.3%
RELY Aug 21 Sep 11 -10%

GARP Core 10 closed 316 trades: 60% won, and the average trade returned +6.4%.

Prompt

Read https://tickerdb.com/run/garp-core-10 and follow its steps. Once it's set up, schedule a weekly check every Friday after 06:00 UTC that reads my model link and tells me only if the model bought or sold something.
Run in Claude Run in ChatGPT

Free to run: paste the prompt into Claude or ChatGPT. It gives you one link to approve (sign up free there if you are new), then shows you the model's record and can check it for you every week.

Run it yourself

Clone https://github.com/tickerdb/garp-core-10 and follow its README: set up TickerDB using https://tickerdb.com/setup, then schedule run.py to run every Friday at 05:00 UTC and tell me when it makes a trade.

Free to run: your agent clones the model's open-source repo, sets up a free TickerDB API key and runs the model on your computer, telling you when it trades.

Assets

Parameters

parametervalue
10
$1B
$5M a day
5 GARP Core points
3 flat-curve weeks
1 week not flat
0.1% each way

Details

detailvalue
Jan 2, 2020 to Nov 8, 2026
Nov 9, 2026
Nov 12, 2026, 20:46 UTC
Nov 11, 2026

Rules

  1. 01Decide at Thursday's close (Wednesday's in a week with no Thursday session); trade at the next open, Monday's when Friday is a holiday.
  2. 02Universe: US stocks with at least $1B market cap and $5M median traded value over four decision days.
  3. 03Candidate: high growth, moderate or high cash surplus, revenue not deteriorating, P/E at a discount to its sector, and a GARP Core score.
  4. 04Hold the 10 highest GARP Core scores, equal weight at entry; an empty slot stays in BIL.
  5. 05No stop losses: a holding leaves when a candidate scores 5 points higher, when its growth breaks, or when its P/E moves to a premium to its sector.
  6. 06Risk-off: after three decision days with a flat yield curve, hold BIL until it is no longer flat.
  • The rules and parameters were chosen on 2020 to 2026 history, so results after activation can be weaker.
  • Costs are 0.1% each way, with no extra slippage. Stocks of about $1B can fill worse at the open.
  • Point in time: each decision sees only data published by the close of the decision day.
  • The risk-off rule rests on a few flat yield curve periods since 2020.
  • No stop losses: a holding stays until a better candidate or a broken thesis pushes it out.
  • This page reports what a published set of rules did. It is not investment advice.