Jolash · Desk Overview
Overview
Where the Jolash project actually stands right now.
- Look at anything marked Promising or with real backtest results and decide: worth chasing, or drop it?
- Pick the paper-trading platform (compare OANDA / Alpaca / MetaTrader demo accounts)
- Write down the exact rules for when we'd move to real money
- Open Hypotheses below, find every yellow "Testing" card
- Copy the command on that card, paste it in your terminal, hit enter
- Report back the real Sharpe/return it prints — even if it's bad news
- Run the research agent on Team & Agents when the hypothesis list is getting thin
- Keep the pipeline/dashboard running and push updates after big changes
- Build out whatever Josh or Francis need next (new features, new data)
Hypotheses tested
29
19 confirmed no edge
Datasets & models
9
on Hugging Face
Agents built
5/5
research + live-data pending
Best backtest Sharpe
1.98
annualized, honest costs included
Held +1000 units long (proba_up 0.575, vol_scale 1.000, drawdown guard untripped).
USDCHF Paper Trading · just now
Full validation on USDCHF 1D h=20: Sharpe 0.44 (2024-25) vs. -0.66 (2021-22) -- REJECTED, failed robustness. Same pattern as the raw carry features.
Central Bank Sentiment Fetcher · just now
Flipped flat -> long (proba_up 0.553 crossed the 0.55 confidence threshold).
USDCHF Paper Trading · 2h ago
Fetched and LLM-scored Fed + SNB statements 2015-2025 -> data/sentiment/USD.csv, CHF.csv.
Central Bank Sentiment Fetcher · 5h ago
First live decision after launch -- position_sizing.py (vol-target scale + drawdown guard) wired in mid-session.
USDCHF Paper Trading · 7h ago
Proposed 4 new combos -- EURUSD 1h h=12, USDCHF 4h h=24, AUDUSD 1D h=12 cleared the cheap screen (NZDUSD 4h h=12 rejected). Second, independent USDCHF signal is notable given USDCHF is already the one confirmed candidate.
Research Agent · 3d ago
Proposed USDCAD 1D horizon=24 and AUDUSD 4h horizon=12 -- USDCAD cleared the cheap screen with a notably strong AUC=0.6223 (small daily sample, needs full validation before trusting).
Research Agent · 6d ago
Proposed EURUSD 4h horizon=12 and USDJPY 1D horizon=24 via OpenRouter (gpt-4o-mini) -- both auto cheap-screened, state persisted to KUTech/trading-llm-agent-state.
Research Agent · 6d ago
Compared fine-tuned vs base model output — fine-tune produced unresolved <think> traces instead of direct answers. Needs dataset rework or retraining.
Jolash Reasoning Model · 7d ago
Ran EURUSD 4h-direction backtest with realistic spread costs — confirmed no exploitable edge (Sharpe -0.86).
Testing Engine · 7d ago
Trained baseline HistGradientBoostingClassifier — validation AUC 0.513, at chance level.
Testing Engine · 7d ago
Completed training run on Colab (Qwen2.5-0.5B, 5000 examples, 157 steps) and pushed adapter to KUTech/trading-llm.
Jolash Reasoning Model · 7d ago