MARKET DATA IN BUY / SELL / HOLD
Live on XAUUSD Gold

An AI that learns to trade

No rules. No indicators. No human bias. OctoBrain watches the market, discovers its own patterns, and trades from experience.

69%Win Rate
+608Pips (Peak)
10,522Patterns Learned
~229Trades / Day
0Lines of Rules

How it works

Imagine a trader who's never seen a chart before. On day one, everything is new. Here's how OctoBrain learns:

Watch & Remember

Every minute, the brain receives 10 numbers describing the current market: momentum, volume, support/resistance proximity, and volatility state. It groups similar snapshots together. Each group becomes a pattern it remembers.

Connect Patterns

When pattern A is frequently followed by pattern B, the brain strengthens that link. When a sequence stops repeating, the link fades. Over time, it builds a web of "what follows what" — like muscle memory for market flow.

BS

Score & Decide

The brain finds the closest stored patterns to right now, walks its web of connections, and produces two scores: BUY and SELL. If one is clearly stronger, it trades. If neither stands out, it waits.

Learn from Results

After every trade, the brain gets a grade. It strengthens patterns that led to winners and weakens losers. It also learns during the trade — updating every minute based on how the position is performing.

No retraining needed No fixed structure No hardcoded rules Adapts in real time Grows its own architecture Learns from every outcome

"An apprentice trader with perfect memory, no ego, and the ability to process 10 dimensions of market data every minute — learning from every single outcome, 24 hours a day."

10,000+ prototypes — each one a strategy the brain discovered

A traditional system might have 3–5 strategies someone wrote by hand. OctoBrain has over 10,000 prototypes — and nobody wrote any of them.

Each prototype is a fingerprint of what the market looked like at a specific moment: momentum was doing X, volume was doing Y, price was this far from support, volatility was expanding… that exact combination becomes a stored pattern.

PROTOTYPE NETWORK — EACH NODE A LEARNED PATTERN

When the brain sees the market right now, it asks: "Which of my 10,000+ stored patterns does this look most like?" It finds the closest match, checks its track record, and acts.

What prototypes capture

Learned to SELL

Momentum fading into heavy resistance with volume drying up

Learned to BUY

Strong volume surge breaking above weak resistance in expanding volatility

Learned to HOLD

Choppy price action, no momentum, near support, low volume

Patterns that stop working naturally fade. Patterns that keep working get reinforced. The brain IS the strategy — there's no strategy file, no rule sheet. Just 10,000+ learned conditions, constantly evolving.

Growth: 500 → 10,522 → 500,000 capacity

500
V1 — Capped
10,522
V2.5 — Current
500,000
Capacity

V1 forced everything into 500 buckets — it couldn't tell a breakout from a fakeout. V2.5 removed the cap: the brain grew to 10,522 on its own, with homeostasis keeping it healthy. The architecture supports up to 500,000. The brain is still early in its learning curve.

What makes this predictive: transitions

Patterns alone aren't enough. Knowing "this is a breakout" is useless unless you know what tends to happen next.

This is where OctoBrain goes beyond what any human trader can do. The brain doesn't just store 10,000+ patterns — it stores 77,222 transitions between them. Every time pattern A is followed by pattern B, that link gets recorded and strengthened.

Think of it like a chess player

A beginner sees the board — that's a pattern. A grandmaster sees the board and the next 5 moves — that's transitions. OctoBrain tracks transitions up to 5 levels deep, simultaneously, across 10,000+ market states. No human can hold that in their head.

Delta
The transition
Pattern A → Pattern B. "After this type of momentum surge near resistance, the market tends to move into this type of pullback." A human trader can track maybe a handful of these intuitively. The brain tracks 77,222 of them with exact probabilities.
Delta²
The transition of transitions
A → B → C. "After that pullback, if volume surges again, the next move tends to be a continuation — not a reversal." This is where human intuition breaks down. The brain sees chains of cause and effect that are invisible to us.
Delta³–&sup5;
Deep sequence memory
"The last 5 market states, in this exact sequence, have historically led to a specific outcome 73% of the time." The brain checks all orders simultaneously — from 1-step to 5-step — and blends them. It's reading the market like a sentence, not a single word.
10,522
Patterns (states)
77,222
Transitions (deltas)
5
Levels deep

A human trader might say "I've seen this setup before." OctoBrain says "I've seen this exact sequence of 5 market shifts before, and here's what happened next — with probabilities across 77,222 recorded transitions." That's the edge no human can replicate.

Momentum + Volume + Micro S/R

Every experienced trader watches two things: is price moving, and is there volume behind it? A breakout on thin volume is a trap. A breakout on heavy volume is real.

OctoBrain doesn't have these as rules. It receives raw data every minute and learns which combinations lead to winners.

What the Brain ReadsWhat a Trader Would Call It
Price distance from 20-bar average, scaled by volatilityMomentum
Whether momentum is speeding up or slowing downAcceleration
Current volume vs 20-bar averageVolume Surge
Whether the candle closed up or downBar Sentiment
Distance to nearest support and resistanceS/R Proximity
Volume traded at that S/R levelLevel Quality
Volatility expanding or contracting vs 100-bar normVol Regime
Whether volatility itself is acceleratingVol Momentum

The vol regime breakthrough

Before (8 features)

Without volatility context

  • Average P&L-2,040 pips
  • ProblemCan't tell real vs fake breakouts
After (10 features)

With volatility regime

  • Average P&L+1,890 pips
  • Gain+3,930 pip swing

Support & resistance — micro level

OctoBrain doesn't use round numbers or daily pivots. It detects S/R in real time from fractal pivot points on 1-minute candles — the actual turning points where price reversed.

Critically, it weights levels by volume. A bounce on light volume is noise. A bounce on heavy institutional volume three times is a wall. The brain learns to follow breakouts through weak levels and fade pushes into volume-heavy walls — without anyone programming those rules.

v1v2v2.5

3 versions in 2 weeks

Version 1 — "The Rules"

March 4–9 • 5.5 trading days

3,600 lines of code telling the brain what to do. 2,200 lines were just rules: when to enter, how to exit, how to size. The brain was one input among many.

The Wipeout

$30.04 → $0.58

Caught an early $14 gold move (+47% in 3 hours). Then the rules took over — overtrading, chasing, compounding losses. By day 5, 98% of the account was gone.

Version 2 — "Let the Brain Decide"

March 11–13 • Architecture reset

Deleted 2,200 lines of rules. One brain, two exits, fixed lot. The brain's own judgment became the only signal. Backtested: best run +9,240 pips.

Version 2.5 — "Brain + Safety Net"

March 13–14 • Current

Added a mathematical safety layer (CbD) that protects the brain's memory from merging patterns that look similar but behave differently in different market conditions.

"The more rules we added, the worse it performed. The less we tell the brain what to do, the better it trades."

CbD: Protecting the brain's memory

As the brain learns, it constantly cleans up after itself — merging patterns that look similar to save space and stay sharp. But sometimes two patterns look nearly identical yet behave completely differently depending on the market context.

Pattern A

Bullish momentum, 0.5 ATR from resistance, volume fading

Context: Calm market → Exhaustion. SELL.
Pattern B

Bullish momentum, 0.5 ATR from resistance, volume fading

Context: Volatile market → Breakout loading. BUY.

Same numbers. Opposite trades. Without protection, the brain would merge these into one pattern and lose the ability to distinguish them.

CbD (Contextuality-by-Default) is a mathematical test borrowed from quantum probability theory. It monitors 9 different market contexts (3 volatility levels × 3 momentum states) and tests whether the relationships between patterns genuinely change across contexts, or just shift in value.

What it prevents
Merging patterns that look similar on paper but trade differently in different volatility regimes. Without CbD, the brain slowly forgets the difference between a breakout and a fakeout.
What it upholds
The brain's ability to discriminate between market regimes. Context-sensitive patterns stay protected. The brain keeps its hard-won knowledge instead of accidentally erasing it during cleanup.
!
What it found
After 79,000+ observations, CbD identified that 33% of context pairs are constitutive — meaning the patterns genuinely restructure between those contexts. The other 67% the brain already handles correctly on its own. CbD is a safety net, not a crutch.

In trading terms: CbD is like having an auditor who watches the brain's housekeeping and says "don't merge those two — they look the same but they trade differently in volatile markets." It lets the brain clean up freely while protecting the patterns that matter most.

v1v2

V1 vs V2.5

V1 — Rules

We told the brain what to do

  • Trades/day~401
  • Win rateUnknown
  • Lines of rules~2,200
  • Lot sizingRisk-based 5%
  • Result-98% wipeout
V2.5 — Brain

The brain decides everything

  • Trades/day~229
  • Win rate69%
  • Lines of rules0
  • Lot sizingFixed 0.10
  • Result+608 pips

March 13–14, 2026

XAUUSD Gold • PUPrime Live • 0.10 lot fixed • 6.8 hours

36
Wins
16
Losses
69.2%
Win Rate
+608
Pips

Session breakdown

19:00–20:35
+214 pips
20:35–21:30
-50 pips
21:30–23:55
+394 pips
00:00–00:15
+50 pips

Biggest winners

PipsDirectionWhat happened
+211.5BUYHeld through $5,094→$5,115 spike (12 min)
+133.2SELLRode $5,051→$5,038 drop, added to position
+130.2SELLCaught momentum $5,060→$5,048
+129.6SELLContinuation trade, full conviction hold

What the brain did

Gold rallied from $5,090 to $5,119 — the brain caught it with BUY signals. When momentum reversed, it flipped to SELL and rode the drop from $5,115 down to $5,030. During the choppy transition between trends, it recognized the chop and reduced its trading. No rule told it to do this. It learned.

0.10

Why fixed lots

We discovered that risk-based sizing (5% per trade) creates a death spiral with tight stops:

+50,310
Pips of edge generated
-$4,825
What risk-based sizing made of it

Each loss compounds, shrinking the next position. Each win grows it — but losses compound faster. Fixed 0.10 lot until we have 500+ live trades to calculate proper Kelly sizing.

Important Disclaimer

One profitable session does not prove an edge. The v2.5 results come from a single 6.8-hour window on a single day. The same brain that caught a clean sell-off might struggle in a choppy, news-driven session. We need 500 closed trades across at least 2 full trading weeks before we can confirm whether the edge is real.

What's next

1

500+ live trades

Statistical significance across different market conditions. One session isn't proof.

2

Multiple sessions

Validate the brain adapts across trending, ranging, and volatile days.

3

Equity curve tracking

Build drawdown profile and risk metrics for proper portfolio management.

4

Graduate lot sizing

Move from fixed 0.10 to Kelly-based sizing once the edge is confirmed.

5

Scale capital

Transition from cents account to standard lots if the edge holds across weeks.

$

Rebate farming potential

The brain's high trade frequency generates significant volume — and volume generates rebates regardless of trade outcome. Even on a break-even day, the rebates produce positive returns.

~229
Trades / Day
$18
Rebate / Lot
22
Trading Days / Mo

The math

Daily volume = 229 trades × lot size

Daily rebate = daily volume × $18 per lot

Lot Size Daily Volume Daily Rebate Monthly Yearly
0.01 (micro) 2.29 lots $41 $907 $10,882
0.05 11.45 lots $206 $4,534 $54,410
0.10 (current) 22.9 lots $412 $9,068 $108,821
0.50 114.5 lots $2,061 $45,342 $544,104
1.00 (standard) 229 lots $4,122 $90,684 $1,088,208

The key insight

At 0.10 lot, the brain generates ~22.9 lots of daily volume. At $18/lot rebate, that's $412/day or $9,068/month in rebates alone — before counting any trading profit. Even if the brain breaks even on P&L, the volume it generates is independently profitable through rebates.

* Assumes 229 trades/day (projected from first 6.8-hour session), 22 trading days/month. Actual volume will vary with market conditions. Rebate rates vary by broker and IB agreement. Pyramiding addons (0.05 lot each) would add additional volume not shown above.

Graduated scaling: trading P&L + rebates

The 69% win rate is from the brain's first live session — with only 10,522 prototypes out of a 500,000 capacity. As the brain accumulates more market experience, discovers finer patterns, and builds deeper transition chains, the win rate has room to climb toward 80%+.

Here's what a graduated scaling plan looks like — increasing lot size monthly as the brain proves itself, combining trading P&L with rebate income:

Win rate trajectory

Month 1
69%
Month 2
72%
Month 3
75%
Month 4
77%
Month 5
79%
Month 6
80%+

Why does it improve? More prototypes = finer pattern discrimination. More transitions = deeper sequence memory. More market regimes observed = fewer surprises. The brain at 50,000 prototypes sees the market with 5× the resolution it has today.

6-month graduated scaling

Conservative estimate: 150 net pips/day base (well below the 608-pip peak), increasing with win rate. XAUUSD pip value: $1 per pip at 0.10 lot.

Month Lot WR Net Pips/Day Trading P&L Rebate Total / Mo
1 0.10 69% 150 $3,300 $9,068 $12,368
2 0.15 72% 180 $5,940 $13,603 $19,543
3 0.20 75% 220 $9,680 $18,137 $27,817
4 0.30 77% 260 $17,160 $27,205 $44,365
5 0.40 79% 300 $26,400 $36,274 $62,674
6 0.50 80%+ 350 $38,500 $45,342 $83,842
6-Month Cumulative
$250,609
Trading P&L + Rebates
Month 6 Run Rate
$1M+
Annualized

Why the numbers compound

Three forces work together:

1. The brain gets smarter. More prototypes, more transitions, higher win rate. 69% is day one — with 500,000 prototype capacity, it's barely started learning.

2. Lot size scales with confidence. Each month's lot increase is earned by the previous month's proven performance. No blind scaling.

3. Rebates scale with volume. Every trade generates rebate regardless of outcome. Higher lots = proportionally higher rebates. The rebate alone at 0.50 lot is $45K/month.

* Projections based on conservative 150 pips/day base (observed peak: 608 pips/day). Net pips/day increases with win rate improvement. Pip value: $1 per pip at 0.10 lot on XAUUSD standard account. 22 trading days/month. Actual results will vary — these projections assume the brain's edge holds and improves as prototype count grows. Lot scaling is manual and conditional on sustained performance.

Bottom line

  • V1 tried to tell the brain what to do. Lost 98% of the account.
  • V2.5 lets the brain decide. First session: +608 pips, 69% win rate, 0.10 lot.
  • The brain fuses momentum, volume, and micro S/R into 10,000+ learned patterns.
  • We need more data before scaling. The brain is learning — the question is whether it has learned enough.