GitHub: https://github.com/abrownfox0/abrownfox001-twap60-prediction-trigger-system
YouTube walkthrough: https://www.youtube.com/watch?v=XzhugRL6BV4
Part 2 inferred the target: live 60s Chainlink TWAP, not CEX last.
Part 3 is the layer nobody can download. You cannot see the weights. You can still infer the constraints any honest P(up) would have to satisfy.
Live profile: @abrownfox001
1. You Do Not Need the Formula
Outsiders waste time asking for the model.
The public record already bounds it:
| Constraint visible from outside | What it implies about P(up)
|
|---|---|
| ~90.6% of fills in 45–55¢ | The signal is used in the coin-flip band, not as an excuse to pay 70¢ |
| Median buy ~47s, not t=0 | The model waits for a readable path, not the opening print |
| Many same-market sells |
P(up) is re-scored; it can decay |
| BTC 5m concentration | The calibration is clock-specific |
Repo splits twapFeed / cexFeed / signal
|
CEX is auxiliary on purpose |
If a bot violates those constraints, it is not this design — even if it copies file names.
2. The Only Interface That Matters
The public hook is simple:
P(up)_t = f(TWAP path vs twap_open, lead-lag, flow, t_left)
Four families. One output. One job: mean what it says.
If 0.57 does not mean “about 57% that official TWAP beats open,” later execution logic is just a faster way to be wrong.
The formula stays private. The interface does not need to.
3. Constraints You Can Infer Without Weights
Constraint A — Official path first
Direction has to come from TWAP versus open.
CEX mid can inform timing. It cannot be the settlement identity.
Constraint B — Threshold, don’t twitch
A mid-band specialist that bought every 50.3¢ flicker would not look like ~40 trades/day over a long sample. The live read is gated.
if abs(P(up) - 0.5) < edge:
wait
if ask not in 45–55¢:
wait
Constraint C — Time is a feature
47-second entries and 226-second exits only make sense if t_left changes the meaning of the same raw probability.
A 52% read at t=20s is not a 52% read at t=280s.
Constraint D — Decay is allowed
~27% sell-side flow is the public proof that the signal is allowed to change its mind.
A frozen-at-entry model would not need that many flat exits.
4. What the Private Layer Is Protecting
Not the CLOB wrapper. Not market discovery. Those are in src/.
The scarce object is a 5-minute probability that stays calibrated after:
- the snapshot era
- the 30s TWAP era
- the current 60s TWAP era
That is why the repo can ship signal.ts as an interface and still say the live read is closed. Publishing weights would compete the only part that is hard to copy.
5. How an Outsider Falsifies the Signal Story
You cannot see features. You can still catch a fake:
- fills cluster at 94–99¢ → not a mid-band forecast
- almost no sells → no live re-score
- size dumped at t=0 every slot → no path wait
- late-second bursts that track one CEX wick → snapshot leftover
- 12 assets, 4 durations → not a calibrated 5m specialist
The public account should keep failing those falsifiers. If it stops failing them, the story changed.
6. The Clean Inference After Three Posts
specialist BTC 5m
+ TWAP60 as the paid variable
+ thresholded P(up)
+ CEX only as lead-lag
+ time-in-slot calibration
+ permission to decay
That is still not the formula.
It is enough to know what kind of machine you are looking at.
7. What Part 4 Covers
Once the signal is allowed to change, the next public evidence is mechanical:
- staggered clips
- re-score cadence
- scratch versus redeem
- why buy timing and sell timing split
The lifecycle is the part an outsider can almost reconstruct from timestamps alone.
Links
- Repo: https://github.com/abrownfox0/abrownfox001-twap60-prediction-trigger-system
- Video: https://www.youtube.com/watch?v=XzhugRL6BV4
- Live activity: https://polymarket.com/@abrownfox001?tab=activity
Not financial advice. Constraints are not a published model.
If you have more questions, please feel free to contact me at any time: https://t.me/abrownfox001
My Polymarket Activity: https://polymarket.com/@abrownfox001?tab=activity
#Polymarket #TWAP #BTC #TradingBot #AlgoTrading #PredictionMarkets #QuantTrading
Top comments (0)