- Add SuspensionFreezeDetector via the IAnomalyDetector seam: a suspension gap with
the favourite unchanged and a negligible (< threshold) price move — the mirror of
the flip. Score = how completely the line froze. Reuses MatchWinEvidence so UI +
evaluator handle it unchanged. 6 tests.
- Add AnomalyKind.SuspensionFreeze + localized card/detail label, SuspensionFreezeThreshold
option, and fan it into DetectAnomaliesUseCase.
- INotificationSink + AnomalyNotification (Application) and a testable
GetPendingAnomalyNotificationsUseCase (date+score filter, event-title join,
oldest-first). 4 use-case tests.
- TelegramNotificationSink posts to the Bot API via HttpClient (no SDK dependency);
no-ops with a warning when unconfigured and never logs the token.
- AnomalyNotificationDispatcher BackgroundService: startup-baselined marker advanced
past the newest sent (gap- and dup-free); idles until Notifications:Enabled.
- Wire options + named client + sink + dispatcher in InfrastructureModule. Add a
secret-free Notifications section + steam-move tuning to appsettings.json
(bot token + chat id go in appsettings.Local.json only).
- RunBacktestUseCase gains an ExecuteAsync(strategy, DateRange?, ct) overload that
pushes the date filter to SQL via IAnomalyRepository.ListByDateRangeAsync; the
existing no-range overload is preserved. +1 use-case test.
- BacktestForm carries optional From/To (Moscow dates) with From<=To validation and
a ToDateRange() helper; BacktestService threads it through. Backtest page gains two
clearable date pickers (empty = all anomalies).
- Localization (en+ru) for the backtest date fields and the settings-validation keys
(shared resx).
- Introduce IAnomalyDetector; the existing flip detector implements it.
- Extract MatchWinEvidence so every detector writes the identical pre/post
evidence shape — the UI parser and outcome evaluator handle new kinds with no
branching (steam moves get hit-rate calibrated for free).
- Add SteamMoveDetector: flags a rapid one-directional implied-probability rise
over a short CONTINUOUS window (no suspension gap inside it), so it never
double-flags the same interval as the suspension-flip detector.
- DetectAnomaliesUseCase fans out over both detectors; dedup keys on EventId+Kind
so flip and steam signals persist independently. Add AnomalyKind.SteamMove +
SteamMove window/threshold options. 8 detector tests.
- Add IEventRepository/IResultRepository.GetManyAsync to kill N+1 lookups at
6 sites (backtest, outcome eval, both bet-journal paths, anomaly browsing,
results selection); guarded by a Received(1).GetManyAsync test.
- Add EventRepository.QueryAsync to push date+sport filtering to SQL (was
load-whole-range-then-filter); search/sort stay in-memory for Cyrillic order.
- Add AnomalyRepository.CountSinceAsync (unread badge) + ListByDateRangeAsync
(feed date filter); add Event/Snapshot count methods for the dashboard.
- Add composite indexes IX_Snapshots_EventCode_CapturedAt and
_EventCode_Source_CapturedAt via a new migration + model snapshot.
- Introduce SqliteDateText as the single source of the O-format date encoding
shared by Mapping (read/write) and the repositories' range predicates.
- Fix LiveOddsPoller cadence drift (budget sleep against cycle time); make
DetectAnomalies dedup O(1) per event; add Event.Title to dedup the title join.
Tests adapted to the batched GetManyAsync via a TestFixtures bridge.
Adds an interactive backtester that replays the SuspensionFlip detector over
all flagged anomalies under a chosen score threshold and staking rule
(flat / percent-of-bankroll / Kelly), and reports the headline numbers a
user needs to judge edge: final bankroll, ROI, max drawdown (peak-to-trough),
win/loss streaks, plus per-bet equity curve.
Domain (pure):
- StakeRule enum + BacktestStrategy params (with validation).
- BacktestSimulator: deterministic function taking strategy + chronological
candidates → BacktestResult. Implements Kelly with post-flip implied prob
as p (skipping negative-edge bets), peak-to-trough drawdown tracking, and
win/loss streak rollups. Mirrors AnomalyOutcomeEvaluator on the 2-way Draw
guard so tennis data inconsistencies are refused rather than miss-counted.
- Skipped counter split into SkippedByThreshold / SkippedByDataQuality /
SkippedByBankroll so the UI can distinguish "strategy choice" from
"data-quality" from "bankroll empty".
Application:
- RunBacktestUseCase: loads anomalies + events + results, parses evidence,
builds candidates, hands event titles into the simulator so the UI does
zero repository round-trips of its own.
UI:
- Pages/Anomalies/Backtest.razor: hero, strategy form (MudBlazor — conditional
sub-field per staking rule), 4-card KPI strip (final bankroll / net profit
/ ROI / max drawdown), counters row, inline-SVG equity curve, trade-trace
table with per-bet outcome pills and link-back to the source anomaly.
- Nav entry under Analysis. RU + EN i18n.
Tests: +20 (16 simulator math — flat / percent compounding / Kelly +/-
edge / quarter-Kelly / bankroll-exceeded / out-of-order chronology / Draw
favourite / multi-window drawdown / event-title pass-through + 4 use-case
join). All 399 tests pass.
Money rounding switched to MidpointRounding.AwayFromZero throughout the
simulator output for accounting convention.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Adds a manual bet-tracking journal that turns the analyzer into an actual
bet tracker. Users record wagers; the journal auto-grades them when event
results land and computes per-bet Closing-Line-Value against the latest
pre-match snapshot — the strongest long-run indicator of betting skill.
Domain:
- PlacedBet entity (reuses Bet vocabulary for Scope/Type/Side/Value/Rate)
with stake, placed-at, outcome, and notes. Derived GrossReturn / NetProfit.
- BetOutcome enum (Pending / Won / Lost / Void).
- BetOutcomeResolver: pure function grading any Match-scope bet against an
EventResult. Handles 1X2, draws, handicap (incl. push), and totals.
Period-scope bets stay manual since EventResult only carries full-time.
Application:
- IPlacedBetRepository abstraction.
- ClosingLineValueCalculator: pure CLV math (implied-probability delta) +
snapshot-matching predicate by Scope/Type/Side/Value.
- BetJournalReport + BetJournalStats records.
- Four use cases: Record / ResolvePending / BuildReport / Delete.
- New ISnapshotRepository.GetLatestPreMatchAsync pushes the closing-line
pick into a single SQLite query rather than materialising the 30-day
window in memory per event.
- ROI turnover excludes Void stakes — pushes are not real turnover and
including them would dilute the user's edge.
Infrastructure:
- PlacedBetEntity / Configuration / Repository / Mapping helpers.
- 20260516 migration adding the PlacedBets table with EventCode and
Outcome indices. Intentionally NO foreign key to Events — the journal
is user data and must survive snapshot-retention pruning. Covered by an
explicit round-trip test.
UI:
- Pages/MyBets/Journal.razor: hero header, 4-card KPI strip (ROI / strike
rate / avg CLV / net profit, tinted by tone), inline add-bet form with
the same invariants as the Bet record, drill-down table with per-row
outcome pills, CLV percentage-points column, P&L, notes underline, and
inline-confirm delete. RU + EN i18n.
- Nav entry under Analysis.
Tests: +55 across Domain / Application / Infrastructure (resolver math
including handicap push and total push boundaries, PlacedBet invariants
and derived properties, CLV math + null-handling, four use cases under
NSubstitute, EF round-trip including survives-event-deletion). All 379
tests pass.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Adds a calibration dashboard that joins persisted SuspensionFlip anomalies
with EventResult rows and reports whether the post-flip favourite actually
won — the single metric that says whether the detector is doing its job.
Domain:
- AnomalyEvidenceData + AnomalyEvidenceParser to read the JSON written by
AnomalyDetector without re-implementing the schema.
- AnomalyOutcomeEvaluator: pure function returning Hit / Miss / Unresolved.
Tennis-style two-way markets with a Draw winner are downgraded to
Unresolved rather than silently counted as Miss.
- AnomalySeverityThresholds: shared Low/Medium/High constants so the UI
badge and the report buckets cannot drift.
Application:
- EvaluateAnomalyOutcomesUseCase orchestrates the join + aggregation.
- AnomalyOutcomeReport carries totals, hit rate, three breakdowns
(severity / sport / score bins) and a per-event title lookup so the UI
needs no second pass over IEventRepository.
- Score bins extend below 0.30 automatically when the operator lowers the
detector threshold so the histogram total always equals ResolvedCount.
UI:
- Insights page at /anomalies/insights — hero header, 4-card KPI strip
(hit rate tinted by tone), three breakdown grids with bar visualisation,
drill-down tables for resolved and unresolved anomalies. Honors
prefers-reduced-motion. RU + EN localisation.
- Nav entry under Analysis section + chip button on the Anomaly Feed.
Tests: +42 across Domain + Application (evaluator boundary cases including
tennis two-way and Draw guard, score-bin edges, dynamic floor when
threshold is lowered, event-title pass-through). All 324 tests pass.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Six MEDIUM-tier review items:
* Mapping.cs — DateTimeOffset.Parse now passes CultureInfo.InvariantCulture
+ DateTimeStyles.RoundtripKind so a non-en-US thread culture cannot
corrupt round-tripped ScheduledAt / CapturedAt / DetectedAt / CompletedAt.
Also replaces the magic 0/1 BetScope discriminator with named constants.
* Delete dead Placeholder.cs files in Marathon.Application and
Marathon.Infrastructure — they were stubs from Phase 1 to satisfy
"non-empty project" and have been dead since Phase 2/3.
* EventBrowsingService — drop the bespoke ScopeEqualityComparer; BetScope
is a record hierarchy, .GroupBy uses value equality natively.
* UserAgentRotatorHandler — counter promoted to private static int with
Interlocked.Increment so rotation is round-robin across the process.
HttpClientFactory builds the handler Transient, so the previous instance
field reset to zero on every new client and broke rotation.
* EventOddsParser — added a parallel "selection-key → IElement" index
alongside the existing price index. Handicap extraction (6 call sites
per event detail page) used to do a fresh document.QuerySelector("span[
data-selection-key='...']") for every key — full-document CSS traversal.
Now it's a dictionary lookup, with the pair-emit logic factored into a
shared TryEmitHandicapPair helper.
* New Marathon.Domain.ValueObjects.MoscowTime with Offset, Now, and
EndOfMoscowDay(DateOnly) — replaces ~15 inline TimeSpan.FromHours(3)
literals across Domain/Application/Infrastructure/UI.
* New Marathon.UI.Services.SportLabels.Resolve(IStringLocalizer, int) —
replaces 6 near-identical SportLabel switch bodies in EventListShell,
Events/Detail, Anomalies/AnomalyFeed, Results/ResultsList,
Results/ResultsLoader, and AnomalyCard. Single source of truth for the
6/11/22723/43658 sport-code mapping. Pages keep a one-liner wrapper so
the call sites stay terse.
* EventBrowsingService.BuildListAsync issued one snapshot query per event
on every page render — N+1 against SQLite, with each round-trip hauling
the full bet graph via Include(Bets). Replaced with a single
ISnapshotRepository.ListByEventsAsync batch.
* ListKnownSportCodesAsync / ListKnownCountryCodesAsync used to materialise
every Event row to compute Distinct() in memory. Pushed to EF projection
via two new IEventRepository methods (ListDistinctSportCodesAsync,
ListDistinctCountryCodesAsync) implemented as
.Select(...).Distinct().ToListAsync — single SELECT DISTINCT.
The Pull*UseCase implementations issued one HTTP request at a time despite
Scraping:MaxConcurrentRequests=4. With 30–80 live events and ~1s per
fetch, a 5–10s live cadence target was unreachable; cycles overflowed
the configured interval.
* New Marathon.Application.Configuration.ScrapingThrottle bound from the
shared Scraping:* section. Exposes only MaxConcurrentRequests so the
Application layer doesn't pull in the Infrastructure-side ScrapingOptions.
* PullLiveOddsUseCase + PullUpcomingEventsUseCase split into two phases:
- Phase 1 — Parallel.ForEachAsync over the event list with
MaxDegreeOfParallelism = throttle.MaxConcurrentRequests. The scraper's
Polly rate limiter still throttles to RequestsPerSecond underneath
this fan-out, so spikes are smoothed before they hit the bookmaker.
- Phase 2 — sequential foreach over the (Event, Snapshot) tuples
captured in Phase 1, doing event upsert + snapshot insert. EF Core
DbContext is not thread-safe so all DB writes stay on a single thread.
* InfrastructureModule binds ScrapingThrottle alongside AnomalyOptions.
* Failed snapshot scrapes in Phase 1 mean the event row is also NOT
persisted in Phase 2 — previously we'd persist the row even when the
snapshot scrape failed, leaving an orphan event with no odds. Updated
the regression test accordingly.
* Test fixture exposes TestFixtures.Throttle(maxConcurrentRequests=1) for
deterministic sequential test runs.
* One existing NSubstitute setup that chained Arg.Is<>() across two
configurations was rewritten to use a single Arg.Any<>() with inline
branching — chained matchers were leaking and returning wrong results.
DetectAnomaliesUseCase was issuing one ISnapshotRepository.ListByEventAsync
call per event each cycle, with each call rehydrating that event's bets via
Include(s => s.Bets) — O(N) SQLite round-trips and N Include payloads on
every detection cycle.
* Add ISnapshotRepository.ListByEventsAsync(IReadOnlyCollection<EventId>, …)
returning a per-event dictionary; events with no snapshots in range get
Array.Empty<OddsSnapshot>() so the caller doesn't need a presence check.
* Implementation uses a single .Where(s => ids.Contains(s.EventCode))
query and groups in memory.
* DetectAnomaliesUseCase loads the whole batch once before the foreach,
then ProcessEventAsync receives the per-event slice as a parameter.
* Tests updated to stub the new method; per-event-failure test now
exercises an AddAsync throw rather than a snapshot-load throw, since
individual snapshot loads no longer fail per-event.
Snapshots are append-only and identified by the composite (EventId, CapturedAt),
not a surrogate Guid. The previous implementation inherited
IRepository<Guid, OddsSnapshot> and faked the lookup with (long)key.GetHashCode()
on a long auto-increment PK — collision-prone and non-portable. Nobody called
GetAsync(Guid) / DeleteAsync(Guid) anyway.
* ISnapshotRepository no longer extends IRepository<Guid, OddsSnapshot>; it
exposes only the methods snapshots actually have: ListAsync, ListByEventAsync,
AddAsync, SaveChangesAsync.
* SnapshotRepository drops the broken Get/Update/Delete methods.
Implements Phase 8 Amendment 1: marathonbet.by has no public results archive
endpoint, so results must be harvested per-event by re-fetching the event
detail page until eventJsonInfo.matchIsComplete=true.
Backend changes:
* IOddsScraper:
- ScrapeResultsAsync(DateRange) replaced with ScrapeEventResultAsync(Event)
returning a nullable EventResult — null when match still in progress.
- ScrapeEventOddsAsync now takes the full Event (so EventPath drives URL
construction) instead of bare EventId.
- New ScrapeLiveAsync() for the /su/live listing.
* Domain:
- Event gains EventPath (nullable string) — the data-event-path attribute
captured during scraping; required for reliable URL construction.
* Infrastructure:
- New migration 20260506000000_AddEventPath adds the column.
- EventEntity / EventConfiguration / Mapping / model-snapshot updated.
- MarathonbetScraper: new ScrapeLiveAsync + ScrapeEventResultAsync; URL
builder prefers EventPath, falls back to numeric ID for legacy rows.
- EventListingParserBase extracts data-event-path on every listing row.
* Application:
- PullResultsUseCase: branches on selection vs date-range, emits IProgress<
PullResultsProgress>, returns ResultLoadOutcome (Loaded / AlreadyLoaded /
NotYetComplete / Failed); idempotent (skips events whose result already
exists).
- PullLiveOddsUseCase now drives off the live listing (auto-discovers
events that go live without ever appearing in the upcoming list) and
backfills EventPath on legacy rows.
- PullUpcomingEventsUseCase wires EventPath on persisted events.
* Workers: UpcomingEventsPoller updates persistence path accordingly.
* Tests: 17 net-new tests across Application + Infrastructure + Domain;
all 293 still pass.
Phase 7 reviewer (Sonnet, combined backend + frontend) flagged 3 🟡 warnings;
two real fixes here, one tracking:
W1 — DetectAnomaliesUseCase had an undocumented N+1: _anomalyRepo.ListAsync
was called inside ProcessEventAsync, once per event. Hoisted to ExecuteAsync
before the loop and threaded into ProcessEventAsync as a parameter. The
per-event slice happens in-memory now. O(N_events) DB round-trips → 1.
W2 — AnomalyDetector.ExtractMatchWinProbabilities had a dead expression
'(decimal?)null ?? 0m' that always evaluated to 0m. Simplified to
'drawBet is not null ? rawDraw / total : 0m'. The 0m is never surfaced
anyway (PDraw in the return uses the same null guard), so behaviour is
identical.
W3 — PLAN.md row updated with both Phase 7 commit hashes (a6ff368 backend
+ 12208a4 frontend) and review verdict.
Build 0/0, 276 tests still passing.
- AnomalyDetector (pure domain): detects odds-flip pattern from live snapshot
timelines using implied-probability vectors (p=1/rate, normalised), flip score
= max(|p_post−p_pre|), gated by both threshold AND favourite-changed test
- SuspensionInterval record: typed pair of (pre, post) OddsSnapshot bracketing a gap
- AnomalyOptions POCO (Application layer): bound to Anomaly:* config section with
four fields (SuspensionGapSeconds=60, OddsFlipThreshold=0.30, MinSnapshotCount=3,
DetectionIntervalSeconds=60)
- DetectAnomaliesUseCase: iterates all events, loads last-24h live snapshots, runs
detector, persists new anomalies with 1-minute dedup window
- AnomalyDetectionPoller: BackgroundService polling every DetectionIntervalSeconds,
gated by WorkerOptions.AnomalyDetectionEnabled (default true)
- DI wiring: DetectAnomaliesUseCase registered Scoped in ApplicationModule;
AnomalyOptions bound + AnomalyDetectionPoller hosted in InfrastructureModule
- WorkerOptions.AnomalyDetectionEnabled added; appsettings.json updated
- 13 domain tests + 4 application tests; total 245/245 passing (no regression)
Phase 4 reviewer (Sonnet) flagged two 🟡 warnings; both addressed:
1. PullLiveOddsUseCase: removed dead 'liveEvents' assignment that called
ScrapeUpcomingAsync but never read the result. Replaced misleading
comment block with a single TODO(phase-6/8) note pointing to the
ListLiveAsync(cutoff) follow-up.
2. Marathon.UI.Services.WorkerOptions.UpcomingScheduleCron: changed default
from '0 */5 * * * *' (every 5 min) to '0 0 */6 * * *' (every 6 hours)
to match Marathon.Infrastructure.Configuration.WorkerOptions and the
appsettings.json default.
PLAN.md: Phase 4 row updated (review status, commit hash); top-level
checkbox in Phases list ticked.
Build 0/0, all 202 tests still passing.
Creates the 9-project .NET 8 solution (5 src + 4 test) with Marathon.Domain
fully implemented: value objects (SportCode, EventId, OddsRate, OddsValue,
BetScope hierarchy), enums (Side, BetType, OddsSource, AnomalyKind), and
entities (Sport, Country, League, Event, Bet, OddsSnapshot, EventResult,
Anomaly) with all invariants enforced in constructors. 96 domain tests pass
(FluentAssertions + xUnit). Directory.Build.props and Directory.Packages.props
centralise build settings and NuGet versions. Both Marathon.sln and Marathon.slnx
are committed; dotnet build Marathon.sln succeeds with 0 warnings/errors.
alexei.dolgolyov