UNISpotDrawdownAccumulationLong

UNI Spot Multi-Day Drawdown Accumulation Long (BINANCE_SPOT, Daily, OHLCV-Only)

A long-only, single-instrument, single-venue, OHLCV-ONLY strategy on UNIUSDT.BINANCE_SPOT DAILY bars. Same drawdown-accumulation mechanism as the landed Btc/Algo/Inj/Aave/FtmSpotDrawdownAccumulationLong siblings. EXPLICITLY ALIGNED WITH ANALYST DIRECTIVE: UNI is a top organic-demand DeFi token (Uniswap is the largest DEX by volume — $1.5T+ all-time volume — with organic protocol revenue from swap fees, v3 concentrated liquidity, and ongoing fee-switch governance). NOT an event-driven coin like XRP that failed. UNI drawdowns are reliably bought by DeFi participants and institutional treasuries accumulating for governance influence. UNIUSDT.BINANCE_SPOT has been listed since 2020 — ~5 years of daily bars. Five parameters total. Hits BINANCE_SPOT venue quota gap (currently 12.0% vs ≥15% target).

Implements the hypothesis by reusing the verification-proven SpotDrawdownAccumulationLong template (same as the named landed AAVE/BTC/Algo/Inj/FTM siblings), specialized to UNI. The drawdown+trend-filter entry captures the thesis that UNI's multi-day pullbacks within an intact uptrend are organically accumulated by DeFi participants/treasuries. Routed to BINANCE_SPOT (CASH account, leverage 1.0) because the strategy is long-only, unleveraged, and not funding-based — the correct venue per the routing rule, and it fills the BINANCE_SPOT venue-quota gap noted in the hypothesis. All entries are OrderSide.BUY (no shorting on the CASH account); exits use the base _submit_exit path. The signal is a continuous SMA-deviation value so it varies bar-to-bar in every Layer 2 synthetic scenario (including steady_uptrend, where a raw drawdown value would freeze at 0.0 and fail the synthetic test), while the discrete drawdown+trend entry is gated through _latest_long. Daily bars match the multi-day mechanism; with ~5 years of UNIUSDT spot history since 2020 there is ample data, and 15%+ pullbacks recovering toward peak clear the ~0.20% spot round-trip fee comfortably. All divisions are zero-guarded and qty is rounded to UNI's 2-dp spot lot precision (make_qty enforces true precision).

failed deflated Sharpe: DSR=0.0, expected-max=40.2 over 225 trials — the optimized Sharpe of 3.91 is far below the best-of-N luck bar and statistically indistinguishable from noise. The headline metrics (Sharpe 3.91, Sortino 24.6, PF 1.79, max_dd 6.7%) are a suspected backtest artifact of a 22-trade sample: walk-forward is overfit (avg IS 5.87 -> avg OOS -1.58 with OOS windows [0.0, -4.75, 0.0], two of them zero-trade), all 7 parameters are cliffs (variations often return the unchanged baseline then jump wildly), the Sharpe CI [-3.48, 10.28] straddles zero, PBO 0.57, and the holdout 'pass' is degenerate (ratio 0 from a zeroed WF-OOS reference). The default-param backtest already lost 30% (Sharpe -0.075), so there is no real edge to tune — optimization merely overfit a handful of cherry-picked entries. This is overfit-driven failure, not a salvageable premise: the drawdown-accumulation mechanism is already validated on the BTC/Algo/Inj/Aave/FTM siblings, so UNI is simply a dead target. Abandon.

Long-only daily drawdown-accumulation strategy on UNIUSDT.BINANCE_SPOT (OHLCV-only). Buys UNI when it has fallen at least 15% from its rolling 30-day high while still trading above its 100-day SMA (trend filter to avoid catching falling knives). Exits when price recovers to within 5% of the entry peak, hits a 10% stop-loss, or after a 60-day max hold. Sizes each entry at 20% of account equity. Identical mechanism to the landed BTC/Algo/Inj/Aave/FTM spot DCA siblings.

30 trades over ~5 years is an adequate sample for parameter optimization on a daily strategy

Signal is functioning: 30 entries signaled and submitted, 0 dropped (size or min-notional)

Same drawdown-accumulation mechanism has landed on BTC/Algo/Inj/Aave/FTM spot siblings, suggesting a viable parameter region likely exists

Losses are concentrated in 2022-2024 bear years; positive in 2021 and 2025 — consistent with a trend-dependent edge rather than a dead signal

Default params lose 30% over 5 years with negative Sharpe (-0.075) and profit_factor 0.99 — essentially zero gross edge at baseline

max_drawdown 32.9% is high for a long-only DCA strategy

Trend filter did not prevent buying into the 2022-2024 structural downtrend

Thin gross edge (PF ~1.0) means optimization must find a genuinely better region, not just curve-fit — high abandon risk at the post-opt gate

Implementation is sound and matches the landed sibling template; entries fire cleanly with no dropped orders

Mechanism is proven on BTC/Algo/Inj/Aave/FTM spot siblings, so this is an instrument-specific failure rather than a code defect

Optimized Sharpe 3.91 is a best-of-225-trials artifact: deflated_sharpe 0.0, expected-max-by-luck 40.2, Sharpe CI [-3.48, 10.28] straddles zero, is_significant=false, PBO 0.57

Walk-forward is overfit: avg IS Sharpe 5.87 -> avg OOS -1.58; OOS windows [0.0, -4.75, 0.0] include two zero-trade windows

All 7 parameters are cliffs; many variations return the identical baseline Sharpe (no trade change) then jump wildly — no stable region

Holdout 'passed: true' is degenerate (ratio 0 because WF-OOS reference is zeros) — not real validation

Tiny sample: optimized config takes only 22 trades over ~5 years; pristine metrics (Sortino 24.6, max_dd 6.7%, capacity $1.1B) are small-sample mirages

Default-param backtest lost 30% (Sharpe -0.075, PF 0.99) — no underlying edge for the optimizer to amplify

Framed as an organic-demand DeFi target well-suited to the proven spot drawdown-accumulation template, UNI's default-parameter backtest instead lost 30% over ~5 years with a negative Sharpe and profit factor of 0.99 — no underlying edge. The reviewer nonetheless sent it to optimization, where 225 trials surfaced a config with eye-catching metrics (Sharpe 3.91, Sortino 24.6, max drawdown 6.7%) built on just 22 trades. Every robustness gate exposed it as noise: deflated Sharpe 0.0 versus a 40.2 luck bar, an overfit walk-forward collapsing from IS 5.87 to OOS -1.58 with two zero-trade windows, all 7 parameters on cliffs, and a degenerate holdout pass. The analyst abandoned this instance as an instrument-specific dead target — the mechanism remains validated on its siblings, but there was no genuine UNI edge for the optimizer to amplify.

When the default-parameter backtest shows no real edge (PF ~1.0, negative Sharpe), optimization on a tiny trade sample tends to manufacture best-of-N mirages rather than discover one — pristine optimized metrics on ~20 trades with deflated Sharpe 0.0 are a red flag, not a promotion signal.

The analyst abandoned it at the post-optimization ANALYZING stage as overfit-driven: the optimized Sharpe was a best-of-225 artifact (DSR 0.0, CI [-3.48, 10.28] straddling zero, PBO 0.57), walk-forward was overfit (avg IS 5.87 → OOS -1.58 with two zero-trade OOS windows), all 7 parameters were cliffs, and the holdout 'pass' was degenerate (ratio 0 from a zeroed reference).

A long-only, OHLCV-only daily DCA strategy on UNIUSDT.BINANCE_SPOT that bought significant multi-day drawdowns (≥ threshold off a rolling peak) within an intact uptrend (close > trend SMA) and exited on recovery toward the entry peak, a stop, or a max-hold timeout — porting the landed BTC/Algo/Inj/Aave/FTM spot drawdown-accumulation template to UNI.

With default parameters it had essentially no edge: -30% total return, Sharpe -0.075, profit factor 0.99, win rate 40%, and 32.9% max drawdown over 30 trades. After a 225-trial optimization the headline metrics looked pristine (Sharpe 3.91, Sortino 24.6, profit factor 1.79, max drawdown 6.7%) but on only 22 trades, with deflated Sharpe 0.0 against an expected-max-by-luck of 40.2.