Reference/Trade Performance Discussion Guide

Trade Performance Discussion Guide

This page is the business-user discussion guide for the Performance page. It separates what the dashboard can currently measure from the data gaps and assumptions that need business sign-off.

Use this page before reviewing Performance with stakeholders. It is intentionally written as a meeting guide rather than an engineering contract.

Executive Summary

The current Performance page is strongest at answering:

Did Alpha trades line up with the dashboard's available opportunity context at entry?

It is not yet a complete execution-quality, forward-outcome, or realized-P&L report.

The most important caveat is that BTC-linked market context can explain the crypto-volatility backdrop for MSTR/COIN Alpha trades, but it cannot prove that a specific MSTR/COIN option trade had good execution liquidity. Execution-quality liquidity needs traded-contract market data at Alpha execution time.

Regime-Aware Copilot Context

The sponsor-facing strategy for the next phase is captured in Regime-Aware Trade Copilot Strategy. It expands the Performance discussion from post-trade measurement toward pre-trade and restructuring support, while preserving the same data-quality boundaries.

The near-term copilot focus is accident avoidance. The system should help identify trades that may look reasonable under the old spot up / vol down regime but are dangerous if current BTC behavior has shifted toward local spot up / vol up, especially in front-month upside vol. Gordon's practical example was a May upside short-vol position where rolling into a larger higher-strike short call structure could add short vol exactly where the surface may be underpricing upside volatility.

This does not remove the business decisions below. It makes them more important: proxy identity, source authority, strategy intent, lookback windows, model confidence, and estimated-mark policy determine whether a warning or restructuring recommendation is decision-grade or only exploratory.

Use the Source Authority Register as the decision checklist for which execution marks, forward marks, terminal marks, realized P&L sources, model estimates, and proxy states can feed decision-grade outputs.

Use Derivative Identity And Proxy Mapping as the decision checklist for exact listed contracts, adjusted contracts, standard-root fallback, SMA/OTC derivative-equivalent mappings, unsupported contracts, and unmapped rows.

Current Production Read

As of the 2026-05-10 production API check, the current 365d Alpha sample shows:

AreaCurrent readBusiness interpretation
Opportunity context749/751 Alpha rows scoredThe dashboard can classify nearly every loaded Alpha trade against available trade-time context.
Opportunity percentage61.5% supportive among decisive rowsThis is not a grade. It means supportive rows outnumber opposing rows among decisive cases, but the aggregate still needs segmentation by portfolio, underlying, regime, and signal.
Alignment distribution322 supportive, 225 mixed/no-edge, 202 opposing, 2 insufficientThe large mixed bucket means the aggregate book is not actionable on its own. Use portfolio, underlying, regime, and signal filters.
Average confidence48.4%Confidence is still reduced by proxy mapping, unstable stationarity, and source-quality limits.
Execution benchmarking331/751 benchmarkedFill-vs-mid, spread capture, and quote-age analysis are available only for rows with sourced MSTR/COIN contract quotes. Missing rows are not bad execution; they are unsourced benchmarks.
Execution liquidity331 partial, 250 unavailable, 170 not sourcedFull liquidity is still unavailable because contract volume and open interest are not populated.
Forward outcome marks177/751 7d marks available; 1 pending and 573 unavailable7d P&L and hit rate are readable only for sourced marks. Unavailable rows are excluded, not treated as zero.
Position coverage751 executions -> 284 derived positionsPosition lifecycle is useful, but it is not authoritative realized P&L.
Realized P&LNot sourcedRealized P&L should remain unavailable until validated against accounting, PMS, broker, or another authoritative source.

ENG-5169 Reliable Ingestion Planning Inputs

This section captures the open questions needed before the reliable Amberdata ingestion build can be split into implementation tickets under ENG-5169.

Known planning decisions as of 2026-05-11:

TopicPlanning position
Ticket structureKeep ENG-5169 as the umbrella planning ticket. Create smaller child implementation tickets and link them as dependencies.
Initial trade universeCover the full historical set of synced Alpha option trades, not only open or recent trades.
First engineering sliceStart with contract identity plus a coverage probe. Every later slice depends on knowing whether each Alpha contract can be mapped and which provider fields are actually available.
Storage shapeUse separate purpose-built stores for execution quotes, liquidity observations, and forward/terminal marks. Shared request metadata can still be reused across stores.
Raw response retentionDefault to compact request/response metadata, row counts, hashes, error/empty-response details, and bounded raw payload retention for failures or sampled diagnostics. Only retain every raw payload if audit/compliance requires it.
Diagnostics taxonomyData Readiness should expose all relevant reason buckets: unmapped contract, adjusted unsupported, proxy used, provider empty, outside market hours, expired before horizon, not entitled, endpoint lacks field, pending job, retry exhausted.
Reliability definitionThe plan should include all reliability requirements: better sourced coverage where possible, explicit reason attribution for every gap, no long-running interactive backfills, and idempotent resumable jobs with checkpoints.

Recommended child-ticket order:

  1. Contract identity and adjusted-contract mapping.
  2. Amberdata endpoint and empty-response validation.
  3. Coverage probe with reason buckets.
  4. Request planner for market-hours-aware quote windows and provider limits.
  5. Durable job/checkpoint framework.
  6. Execution quote store and benchmark ingestion.
  7. Liquidity ingestion through dedicated volume/OI sources.
  8. Forward and terminal mark ingestion.
  9. Data Readiness diagnostics and operator reporting.

The first slice should be contract identity plus provider coverage probing because it reduces the largest uncertainty before storage, job, and UI work. It also gives the business an auditable gap report before we start filling tables with partial data.

Open Questions For Full Ingestion Plan

These questions should be answered or explicitly marked as engineering-owned before implementation tickets are created.

Business and source-system questions

QuestionWhy it mattersNeeded from
What exact Alpha fields are available for option identity: OCC symbol, OPRA/native symbol, root, expiry, strike, put/call, multiplier, deliverable, exchange, or provider identifier?Determines whether the contract identity layer can map Alpha rows without guesswork.Alpha / OMS owner
What is the authoritative source for adjusted 2MSTR and 2COIN symbology, deliverables, multipliers, and cash-in-lieu treatment?Standard-root fallback is not authoritative adjusted-contract valuation.Alpha, broker/OMS, OCC/OPRA, OptionMetrics, or approved vendor
Where, if anywhere, is standard MSTR / COIN proxy fallback acceptable for adjusted contracts?The policy may differ for diagnostics, execution benchmarks, forward marks, liquidity, and P&L.Business owner
Which source should define terminal or expiry marks when a 1d/7d/30d horizon falls after option expiry?Determines whether expired-before-horizon rows use expiry quote, close mark, settlement mark, broker/PMS mark, or remain unavailable.Business owner / operations
What is the authoritative realized-P&L source?Realized P&L should stay unavailable until it is validated against accounting, PMS, broker, settlement, Alpha output, or lifecycle accounting.Accounting / PMS / operations
Can Alpha provide strategy intent, trade reason, opening/closing flag, or parent-order grouping?Prevents buy/sell direction from being overread as complete strategy intent.Alpha / trading owner
Are model marks, stale quotes, or nearest-session marks acceptable as explicitly labelled estimates?Determines whether estimated states can supplement sourced states, and where estimates must remain excluded.Business owner
What sourced-coverage threshold is acceptable for release sign-off if provider coverage cannot reach 100%?The build can attribute every gap, but business needs to decide the minimum sourced coverage needed before reading the report operationally.Business owner
Is full raw provider-payload retention required for audit, or is compact metadata plus bounded failure/sample payload retention sufficient?Changes storage volume, retention policy, and privacy/audit posture.Compliance / operations if applicable

Provider and engineering validation questions

QuestionWhy it mattersOwner
Which exact Amberdata endpoints support instrument discovery for historical TradFi option contracts?The mapper needs provider-queryable identifiers rather than inferred strings.Engineering / Amberdata
Which exact Amberdata endpoint should provide historical level-1 bid, ask, mid, mark, quote size, and quote timestamp?Execution benchmarks and forward marks depend on field-level semantics.Engineering / Amberdata
Which exact Amberdata endpoints provide contract-level open interest and contract/date volume for MSTR/COIN options?Full execution-liquidity coverage cannot rely only on level-1 rows if OI/volume are absent there.Engineering / Amberdata
Is there a suitable terminal, close, or settlement mark endpoint for expired option contracts?Expired-before-horizon rows cannot be valued with live horizon quotes.Engineering / Amberdata
What does HTTP 200 with empty data mean for each endpoint?Empty data could mean bad instrument, no quote, no entitlement, no historical coverage, outside market hours, or expired contract. Each case needs a different reason bucket.Engineering / Amberdata
What are the practical request limits: historical retention, max time window, max instruments per request, batching support, rate limits, retry behavior, and entitlement boundaries?The request planner and job checkpointing rules must fit real provider constraints.Engineering / Amberdata
Can requests be batched by instrument/time, or must the backfill issue one contract/window request at a time?Drives job duration, rate-limit strategy, and whether Cloud Run request-time execution is viable.Engineering / Amberdata

Data Gaps To Discuss

1. Execution-time option quotes

Execution benchmarking is partial because the dashboard has sourced MSTR/COIN option bid/ask quotes for some Alpha execution rows, but not for the full trade population.

Required fields:

Required inputWhy it matters
Contract bidNeeded to evaluate sell fills and spread.
Contract askNeeded to evaluate buy fills and spread.
Contract midBenchmark for fill-vs-mid.
Bid/ask spreadNeeded for spread capture and liquidity quality.
Quote timestamp / quote ageNeeded to reject stale quotes.
Contract OINeeded for trade-level liquidity context.
Contract volumeNeeded for trade-level liquidity context.
Contract identifier matchNeeded to join Alpha trade rows to the market quote source.

Current provider validation found fresh sampled MSTR/COIN contract quotes with bid, ask, mark, bid/ask size, and quote timestamps. The sampled Amberdata level-1 rows did not include 24h contract volume or open interest. The dashboard therefore treats sourced bid/ask liquidity as partial execution-liquidity evidence until volume/OI are returned by the provider or sourced elsewhere.

Business question:

What source should be authoritative for MSTR/COIN execution-time option quotes?

Possible sources include Alpha, broker/OMS, OPRA, OptionMetrics, Amberdata TradFi coverage if entitled, or another approved market-data source.

2. Forward marks and outcomes

Forward outcomes are partial because the dashboard has sourced option marks for some trades and horizons, but not for the full MSTR/COIN trade population.

Business question:

Which source should define post-trade marks and P&L horizons?

The answer matters because hit rate and average P&L should exclude unavailable marks rather than treating them as zero.

3. Realized P&L authority

The dashboard can derive position lifecycles from Alpha rows, but derived same-contract cashflow should not be presented as realized P&L until validated.

Business question:

What is the authoritative realized-P&L source?

Candidates include accounting, PMS, broker statements, Alpha output, or a validated lifecycle accounting process.

4. Strategy intent

The current opportunity model assumes:

buy = long volatility
sell = short volatility

That is a useful first pass, but it may be wrong for hedges, rolls, spreads, closing trades, directional overlays, and portfolio-level constructions.

Business question:

Can Alpha provide strategy intent, trade reason, opening/closing flag, or parent-order grouping?

Without strategy intent, the dashboard should keep saying that opportunity alignment is assumption-based.

5. BTC proxy context

MSTR and COIN Alpha trades currently map to BTC-linked market context. This is appropriate for crypto-volatility backdrop, but it is still a proxy.

Business questions:

Where is BTC proxy context acceptable?
Where is it not acceptable?
Should MSTR and COIN have separate equity-option context layers?

Current recommendation:

Context typeRecommended treatment
BTC vol regime, DVol, VRP, GEX, stationarityAcceptable as market backdrop if visibly labelled as proxy.
MSTR/COIN execution liquidityMust be sourced from traded-contract option data, not BTC proxy data.
MSTR/COIN smile richnessKeep not sourced until equity-option smile data or an approved proxy method exists.
Realized trade P&LMust come from authoritative trade/accounting data.

6. Context coverage and freshness

Current source-group coverage has narrowed for opportunity context, but the table still matters because it distinguishes market backdrop from execution-quality evidence:

Source groupCurrent 365d stateInterpretation
GEX / Gamma749 available, 2 unavailableBTC proxy regime context is available for nearly all rows.
Stationarity749 available, 2 unavailableStationarity is available for nearly all rows and should be checked before trusting conditional probability context.
Vol Metrics749 available, 2 unavailableDVol/VRP/IV/RV context is now available for nearly all rows in the current production sample.
Term Structure749 available, 2 unavailableTerm-structure context is now available for nearly all rows in the current production sample.
BTC Proxy Liquidity749 partial, 2 unavailableBTC proxy liquidity remains backdrop only. Partial means it is not complete traded-contract execution liquidity.
MSTR/COIN Execution LiquidityPartial where fresh contract bid/ask exists; unavailable/not sourced where provider lookup fails or adjusted contracts cannot be safely mapped.This is the execution-quality liquidity layer. Missing contract volume/OI remains unavailable rather than zero.
Smile751 not sourcedDeliberately excluded for Alpha MSTR/COIN rows.

Business question:

Should the next phase prioritize backfilling context coverage, sourcing execution/outcome data, or adding strategy intent?

Assumption Gaps To Sign Off

AssumptionCurrent behaviorWhy sign-off matters
Buy/sell intentBuys are treated as long-vol; sells as short-vol.This can misclassify hedges, rolls, closes, or directional trades.
BTC proxy mappingMSTR/COIN opportunity context maps to BTC.This is a backdrop assumption, not direct instrument evidence.
Mixed rowsMixed/no-edge rows are visible but excluded from the headline decisive percentage.This changes headline interpretation from an average score to a decisive-context percentage.
ThresholdsDVol, VRP, gamma, stationarity, and term-structure point rules are hard-coded.Business should agree that the thresholds match how they judge opportunity alignment.
Confidence penaltiesProxy mapping, low signal coverage, and unstable/insufficient stationarity reduce confidence.Confidence should communicate trust, not whether a trade was good or bad.
Unavailable dataMissing fields are not counted as zero.Prevents data gaps from being read as poor performance.

Recommended Business Conversation

  1. Start with the Performance page Data Readiness section.
  2. State that opportunity alignment is scored, but execution quality and outcomes are not yet sourced.
  3. Explain the current 61.5% as a strategy-read percentage among decisive rows, not a grade.
  4. Show the large mixed bucket and ask whether the business wants segmentation by portfolio, regime, underlying, or strategy intent.
  5. Explain that BTC proxy liquidity is market backdrop only.
  6. Ask which data source should be authoritative for execution-time MSTR/COIN option quotes.
  7. Ask which data source should be authoritative for forward marks and realized P&L.
  8. Ask how adjusted 2MSTR / 2COIN contracts should be mapped and where standard-root proxy fallback is acceptable.
  9. Ask what sourced-coverage threshold is acceptable for release sign-off if provider coverage cannot reach 100%.
  10. Review the opportunity signal point rules and decide whether thresholds should be accepted, adjusted, or made configurable.

Business Decisions Needed

DecisionNeeded before
Alpha option identity fieldsBuilding the contract identity layer.
Authoritative adjusted-contract mapping and deliverablesTreating 2MSTR / 2COIN as anything other than explicitly caveated proxy coverage.
Adjusted-contract proxy policyUsing standard MSTR / COIN fallback in diagnostics, execution benchmarks, forward marks, liquidity, or P&L.
Authoritative execution quote sourceDisplaying fill-vs-mid, spread capture, or execution quality.
Authoritative forward mark sourceDisplaying 1d/7d/30d P&L, hit rate, or outcome attribution.
Authoritative terminal/expiry mark sourceDisplaying expired-before-horizon outcomes.
Authoritative realized-P&L sourceDisplaying realized P&L or success/failure metrics.
Accepted use of BTC proxy contextPresenting proxy market context in Alpha measurement views.
Strategy-intent sourceMoving beyond buy=long-vol and sell=short-vol assumptions.
Estimated mark policyShowing model, stale, or nearest-session marks as explicitly estimated rather than sourced.
Reliable-ingestion release thresholdShipping the first production version if sourced coverage remains below 100% but every gap has a reason bucket.
Raw provider-payload retention policyDeciding whether compact metadata is enough or full raw payload archival is required.
Opportunity thresholdsTreating supportive/mixed/opposing labels as business-approved.