Revenant Live Dashboard

One chiclet per MLB club. Double-click a team for batting and pitching form: last-10 game bubbles — batters & starters are green over season average, yellow near it, red below; bullpen/closers are green on a save (or clean 0-run outing), red on a blown save or runs allowed.

No teams loaded yet.

From Discord !bet / !bets — daily hit rate for ML, NRFI, under through 5th, and under through 9th. Bars are W / (W+L); pushes and open bets sit in the chips but not the %. Click a day for that slate’s tickets.

Same model as !winprob: expected runs = own R/G × opposing pitching factor (SP 5.5 IP + BP 3.5 IP), home gets +4% HFA, then Pythagorean (k=1.83) → win %. The highlighted side is the pick. Pregame picks lock the first time we compute them; Finals without a lock are graded by replaying the model with that game's actual starters (current season rates). A green border means the pick won; red means it lost. Under each team's win %: last 10 results as bubbles (green = win, red = loss; solid = home, ring = away), starting from that team's last Final (the same game as the linescore box) and working backwards — opponent abbrev beside each bubble and that game's H/R/HR, opposing starter ERA, and the model's pregame win % for the team underneath. Double-click a chiclet to pin both starters' last 5 outings (!pitcher table) plus each team's last 5 Finals batting vs opposing SP vs bullpen (hits / runs / HR / times on base, with the ERAs faced). The chart is the daily hit rate over the last 30 days including today — hover a bar for that day's pick-% bucket W–L (50-55 / >55-60 / >60, color-matched to the chips above); click for the full breakdown (today opens by default when the slate has graded picks). Today's pick-% chips also sit above the chiclet wall next to the today W–L status.

No games on this slate yet.

Today's slate always loads below (Live + Scheduled — filter in the toolbar). Click a history bar to open that day's chiclets in a day card (same idea as 1st-inning); use Today’s games or Clear to return. Double-click a chiclet to pin its !game live-style board — full linescore, batters, pitchers, and each side's Pitcher Card (current pitcher's last 6 appearances). In Last 6 games, amber names also pitched the previous game.

A live gas-tank read on every starter as the 5th approaches, headlined by the run distribution from now through the 5th (P(≤0) … P(≤3), 4+) — the same Next-N model as !scoreless (each remaining half is a negative binomial expecting the current pitcher's ERA ⁄ 9 runs × 1.06 for unearned runs — ERA only counts earned; runs already in this half shift it). Each side shows the pitcher's pitch count, hits+walks, strikeouts, and swing-and-miss rate (SwStr% from the live pitch feed). A green MISS tag lights when SwStr is ≥ 14% on 25+ pitches (league starters sit ~11%). A red GAS tag fires when the starter is still in with his 5th unsettled and any of: pitch count ≥ 80, traffic ≥ 7 H+BB, or 18+ batters faced (the 3rd time through starts) — backtested, the run line through 4 is noise but traffic/fatigue are real. Dots are runs allowed in innings 1–5; a ✓/✗ settles each side's 5th. Only games still in the 5th inning or earlier appear here — once a game hits the 6th it moves to the 9th tab.

No live games in the 5th inning or earlier. Later games show on the 9th tab; this wall polls every 10s while open.

The same read pointed at the 9th: most likely run counts from now through the 9th (P(≤0), P(≤1), P(≤2), P(≤3), 4+), priced off whoever is currently on the mound for each side — the !scoreless Next-N negative-binomial model (ERA ⁄ 9 expected runs per remaining half × 1.06 for unearned runs; a bullpen-average ~4.0 ERA fills in when a pitcher has no season ERA yet). Future halves are priced off the projected bullpen ladder (availability + roles + usage, the !bullpen engine) — in a close game that means the setup man and closer, not whoever is on the mound now; hover the bars to see the projected arms. A ✓/✗ settles each side's 9th. Dots are runs allowed in innings 5–9. The bottom of the 9th is priced on the score: skipped when the home side leads after the top (the game just ends), and truncated at the walk-off run when it is played. Only games already in the 6th inning or later appear here.

No live games in the 6th inning or later. Earlier games show on the 5th tab; this wall polls every 10s while open.

Games that looked like this. No model — pure history. Each live game's partial linescore (which innings the runs came in, plus hits so far — including the open half) is matched against every completed game this season, and the card reports what actually happened in the 100 closest lookalikes: how often each side won, how many more runs scored the rest of the way — and, before the 5th is done, also through the end of the 5th (P(≤0) … P(≤3), 4+ — reads like the 5th/9th tabs) — and how often the side trailing at this point came back. Ranking is tiered: same score at this point first, then which innings the runs came in, then hits, then how close the two starters' season ERAs sit to tonight's; exact = same score, pattern, and hits, close = same score within one run out of place or ~3 hits. A run that just scored in the open half is in the match (so bot 7th 3–2 matches other 3–2 games through bot 7, not the pre-run score). Each card also shows the plain same-score cohort — every season game that was exactly this score at this point, whatever the path, and who went on to win. The closest five matches each show a condensed innings/runs linescore box — innings inside the matched boundary (bot 7 = through the bottom of the 7th so far, top 6 / mid 7 / end 6 likewise) read muted, and the amber innings after it are what happened next (x = a home 9th that was never played). Amber r/h are runs and hits through that matched point; final R/H follow. Under each header: that game's starting pitchers with season ERA and innings pitched that day. Matching starts once two halves are on the board; deeper games match tighter (more innings pinned down), so treat early-inning cards as weak reads. Auto-refreshes about every 30s while this tab is open (or hit Refresh). Use Lock to freeze the wall while you read — Unlock when you want auto-refresh again. Powered by the season history cache — refresh with !history refresh if the pool looks stale.

No live games right now — lookalike cards appear as games go live (once two halves are on the board).

This club scored in these innings vs this kind of ERA. Not a full-path lookalike (that's Similar). Each live side that has scored is matched to that club's other season games with the same scoring pattern so far (quiet innings stay quiet — 0-0 then a 3rd counts), against a starter whose season ERA sits within ±0.50 of tonight's opposing arm. The card then shows what happened the rest of the way: did they score again before the 5th closed? After the 5th? And the actual linescores (focus innings highlighted; amber = after tonight's current inning). Tonight's linescore sits on the card; away lookalikes are on the left, home on the right. Home matches other home games (bottom innings); away matches other away games (top innings). Analog linescores keep away on top. Prefer ERA-window games, ordered by how close the hit line was at the matched inning, then closest ERA; if fewer than three, fill with the nearest ERA. Each lookalike shows hits at the match (h) and at the final (H). One completed half is enough (top of the 1st with a run counts). Auto-refreshes about every 30s while this tab is open. Use Lock to freeze the wall while you read. Historical ERAs are disk-only — refresh the dashboard cache / !history refresh if the pool looks thin.

No live scoring innings yet — cards appear once a club puts a run on the board (from the first completed half).

Today’s slate always shows under the charts (like WinProb). Pick record, SP record, and calibration stay under that wall and load in the background after the chiclets paint — the buttons jump down to each board. Click a yellow daily bar to swap in that day's games ranked by P(0 runs 1st), with a green ✓ when the 1st stayed scoreless and a red ✗ when a run scored; use Today’s games to come back. The subtitle shows that wall's per-game max starter ERA p50 / p75 / p90 (higher of the two starters per game; 0.00 / missing / tiny-sample ERA > 12 excluded). The daily bar chart overlays each day's max-ERA p50 as a blue line on the right-hand axis (scaled to the window's own hi/lo so the movement reads). Under the full-inning % each chiclet also shows T1 / B1 P(0) for the top and bottom halves. Same layout as Today’s games. Home starter on top, away underneath. The through-1st score pill shows each side's runs with a blue matchup edge (opp pitcher power − this lineup's batting power; 0 when the pitcher has no power number yet). Under each: his last ten starts (green = scoreless 1st, red = run(s) allowed; solid = home, shaded ring = away), then the opposing lineup's last 7 venue-matched games (away batting strip = last 7 road 1sts; home batting strip = last 7 home 1sts; same colors / venue shading). The trailing % after a lineup strip is that club's season quiet-1st rate at this venue only (pitcher strips stay overall). Each row also ends with a blue 0–100 power index (50 = average): lineups score higher for 1st-inning runs off good starters; pitchers score higher for blanking strong lineups and lose more for runs allowed to weak ones. Both weight recent games more than April's. A blue border highlights standard scoreless-1st picks (model ≥ 55%, and each half needs its blue edge ≥ −5 or a lineup scoreless in 5+ of its last 6); an amber border with a PRIME tag is the higher bar (model ≥ 60%, non-division) — backtested at ~55% / ~71% vs a 49% base over the last 60 days. A red border with a RUN tag is the opposite call — a run scores in the 1st: model ≤ 48% scoreless and at least one lineup's batting power beats the opposing starter's quiet-1st power by 10+ (edge ≤ −10) — backtested at ~68% vs a 51% base over the same 60 days. Matchups in red with a DIV tag are division games, where the model's high-confidence calls hit only ~58% (so Prime is withheld there). Under each lineup strip, that team's last game shows as hits/runs/HR, in amber when they were blanked (≤1 run or ≤3 hits) — bounce-back watch — with 1–3 hits/HR from that game's top of the order underneath. Hover any chiclet or number for details. Double-click a chiclet to pin the board above the wall — for completed games, the Situation-style full linescore first, then both starters' last 5 start linescores (Pitcher Card P/H/BB/R/HR by inning) vs opposing batting. Use Sort for Percentage (P(0) high→low) or Time (first pitch); that clears a saved drag order. Drag chiclets to reorder (on touchscreens, press and hold a chiclet first — a plain swipe scrolls); the custom order is remembered locally. Star a chiclet to cycle bookmark colors in place (unfilled → green → yellow → grey → red → clear; saved locally; order stays put). Power index opens season-long leaderboards — batting teams on the left, starters (with team abbrev) on the right — both sorted high → low. Computed from the season history cache (per-inning linescores for every completed game). Rebuild it with revenant --history refresh or !history refresh to pull in recent games. Pick record grades the highlighted picks over the last 30 days: the overall win/loss record on top (Prime broken out in amber), then the same pick-% buckets as WinProb (50-55 / >55-60 / >60) grading every settled game by model P(0 runs 1st) (win = scoreless), a bar per day (its wins/picks label above each — hover for that day's color-coded bucket W–L), and every pick chiclet grouped by date below — click a bar to pin that day's pick chiclets (pitcher names included). SP record does the same for the green SP / SP+ / SP★ starter-quality tags: grouped bars are that day's % with a run for all SP tags, SP+ only, and SP★ (lower is better for NRFI; SP+ excludes stars, same as raw SP excludes plus and star). Each day is labelled SP–SP+–SP★ with-run/n (0/0 when a tier had no game) so a clean 0/1 still reads when the bar is flat. Lines are each tier's running NRFI win% from the left of the window, the season base rate is the dashed line, the aggregate NRFI W–L is on top (disjoint SP / SP+ / SP★ broken out), then SP / SP+ / SP★ by model P(0) band versus every settled game in that band (lift). Click a band (season or this window) to pin those chiclets — e.g. season SP >60 9–7. Click a column to pin that day's highest SP chiclets (SP★, then SP+, then SP) under the chart. Calibration checks the model against reality over the same 30 days: every settled game lands in a 5-point bucket by its predicted P(0 runs 1st), and each bucket's bar shows how many actually stayed scoreless (wins/games above each) against the dashed average-predicted line — bars near the line mean the percentages can be trusted at face value. The Model dropdown picks the prediction logic everywhere on this tab: v1 blends league half baselines, starter ERA, and recent offense; v2 additionally grades each lineup on its own venue only (home teams by home 1sts, away teams by road 1sts, min 8 venue games, falling back to v1 otherwise). Displayed %, T1/B1, sort, and the calibration board follow the selection; the blue/amber/red pick borders stay on v1, where they were backtested.