X3 Phase 1 Validation
Avanza MA Strategy replay vs. synthetic X3 model. Created 2026-05-13.
Conclusion
Phase 1 is substantially validated. After accounting for known trade timing from Avanza contract notes, the synthetic replay lands very close to the real Avanza account value at the latest checkpoint: +0.84%.
The large initial mismatch was not a fundamental X3-model/accounting failure. It was mostly caused by comparing real Swedish morning / intraday certificate trades against a daily close-to-close TSLA model.
What We Tried To Validate
The purpose was to test whether the model can explain Yusuf's real Avanza MA Strategy account when given the same real-world account activity: deposits, withdrawals, buys, sells, tax, interest, risk costs, and trade dates.
This validates the X3 component used by the strategy report: synthetic X3 pricing, USDSEK handling, daily reset math, transaction accounting, and cashflow handling. It does not validate IBKR option fills, collar spreads, put pricing, or assignment mechanics.
Inputs
- Private local Avanza transaction export:
data/avanza_ma_strategy_transactions.csv - Manual Avanza NAV checkpoints:
analysis/validation/avanza_ma_strategy_nav_points.csv - Extracted contract-note timing sample:
analysis/validation/avanza_ma_strategy_trade_notes_sample.csv - Replay script:
scripts/replay_avanza_ma_strategy.py - Generated comparison output:
analysis/validation/avanza_ma_strategy_nav_comparison.csv
The raw transaction export and PDF contract notes contain personal account details, so the public report links only the sanitized extracted artifacts where practical.
Method
We started the replay at 2024-04-23, immediately after the account appears to be cleanly liquidated from earlier Tesla shares and mixed certificates. From that cash-only point, the replay applies later Avanza account rows and marks positions using the local rate-aware synthetic X3 model.
The first date-only replay was much too high versus Avanza. We then extracted execution times from selected Avanza avräkningsnotor. Those showed that many large mismatch trades were executed in Swedish morning, before the US regular session opened. For those known pre-US-open trades, the replay uses the previous US trading close as the daily-data proxy. Known regular-session trades remain same-day close proxies.
Results
| Date | Avanza NAV | Model NAV | Diff | Diff % |
|---|---|---|---|---|
| 2024-12-31 | 126,162 kr | 114,314 kr | -11,848 kr | -9.39% |
| 2025-12-31 | 140,102 kr | 146,335 kr | +6,233 kr | +4.45% |
| 2026-05-10 | 112,876 kr | 113,823 kr | +947 kr | +0.84% |
Avräkningsnotor Explanation
The contract notes explained the initial scary gap. For example, several sampled October 2025 trades were made at Swedish times like 08:32, 09:09, or 10:20. During that period the US regular market opened around 15:30 Swedish time, so those fills were based on issuer/premarket certificate pricing, not TSLA's US close.
After adding note-based timing overrides, the latest checkpoint moved from a large miss to a close match. This is strong evidence that the X3 model/accounting is broadly right, while exact trade replay requires intraday or premarket execution data.
Limitations
- The replay uses daily TSLA/FX/rate data, not tick-level certificate pricing.
- Some remaining differences are expected because real certificate fills can happen intraday, premarket, or near US open/close.
- The 2024 checkpoint is weaker because that period includes more mixed instruments and less complete note timing.
- This validates the X3 sleeve/accounting, not the option/collar execution layer.