Pirates at Braves Sunday: Highest XGBoost Signal on Today's MLB Board

XGBoost scores 58, composite confirms at 58, and the h2h line moved 84 pips toward Atlanta. Here is what every market shows before the rubber match.

GhostLeg, 2026-06-07, 8 min

The highest XGBoost signal on today's MLB board

Thirty-six minutes past 1 p.m. ET on Sunday, the Atlanta Braves and Pittsburgh Pirates take the field for the rubber match of a three-game weekend series. The Braves are already up two games to nothing, both victories by an identical 6-3 score. They are one win away from a sweep at Truist Park.

That coincidence, two games at the same 6-3 final, is a footnote. The more interesting number arrived from GhostLeg's signal engine this morning: a 58 XGBoost edge score on the Atlanta h2h. That is the highest XGBoost reading among all upcoming MLB games on today's board.

The composite score, GhostLeg's multi-signal rollup, lands at 58 as well. Both signals point the same direction: the Braves at home, going for the clean sweep.

And this time the line agrees with them. Before getting to the movement, it is worth understanding how the XGBoost layer works and why this game surfaced at the top.

What the XGBoost model layer is

GhostLeg's signal pipeline runs several model layers in parallel. The composite score (0-100) aggregates a wide basket of signals: line movement, sharp-money consensus, prediction-market alignment, historical context. It is the number most bettors see first in the intel dashboard.

The XGBoost layer is newer. It is a gradient-boosted decision tree trained on graded historical bet data, evaluating the gap between the price currently on offer and the model's estimate of fair value. The output is also a 0-100 integer, where higher scores indicate more model-perceived edge. An XGBoost score of 58 on a specific game does not mean the game is a lock, and it does not predict a win rate. It means the model, looking at the offered prices against its trained sense of fair value, sees more edge on this game than on any other MLB contest on the board today.

When XGB (58) and composite (58) land on the same number and point the same direction, that internal coherence adds weight. Two different evaluation frameworks, built on overlapping but distinct feature sets, arriving at the same reading. Follow both live through the GhostLeg intel feed.

What the books say

Seven sportsbooks priced the Pirates-Braves h2h for Sunday. The closing lines:

Median sits at ATL -146 / PIT +124. Pinnacle, the sharpest book in the market, is the most generous toward the underdog at Pirates +129. BetOnline also shows +126 on Pittsburgh. If you are looking at the Braves side, BetMGM and DraftKings tighten their Atlanta price most, at -150 and -149. The spread in favorite pricing across books is narrower here than you see in marquee playoff basketball, reflecting a relatively settled market.

At -146, the implied probability for Atlanta on the consensus price is roughly 59.3%.

The run line and the total

The MLB run line sits at Atlanta Braves -1.5 across all seven books. That is baseball's standard spread. The price for taking Atlanta to win by two or more runs lands in the +135 to +142 range, with Pinnacle offering the best number for Braves run-line backers at +142. Going the other direction with Pittsburgh +1.5 costs between -155 and -166, with BetOnline showing the friendliest price on the Pirates side at -155.

Run-line composite score is 54, a moderate signal but weaker than the h2h read at 58.

The total shows a half-point split across books. Most books (BetMGM, Bovada, DraftKings, Fanatics, FanDuel) post the line at 8.5. BetOnline and Pinnacle sit at 9.0. Pinnacle's Over 9 is priced at +104; most books post Over 8.5 in the -118 to -120 range. That half-point matters for push scenarios, particularly if this game follows the 6-3 template of the first two. Both G1 and G2 landed at nine total runs exactly.

Totals composite score is 56 with the Over as the signal pick.

What the prediction markets say

Three venues. Here is what each priced before game time.

Pinnacle implied: 58.8% Atlanta (pin_implied). This is Pinnacle's market-derived win probability for the home team after removing the vig.

Kalshi: 58% on Atlanta. Kalshi carries all three markets for this game: 2 h2h contracts with $70,720 in volume, 6 spread contracts at $6,693, and 11 totals contracts at $3,283. Volume is modest compared to what you see on Kalshi during NBA playoffs, which is normal for regular-season MLB. All three markets are live, and the h2h probability aligns tightly with the sportsbook consensus.

Polymarket: 56.5% on Atlanta. Polymarket has a head-to-head market with $2,381 in volume. At that depth, it functions as confirmation rather than a standalone directional signal. The number is in the same band as Kalshi and Pinnacle, so it holds.

GhostLeg's pre-game intelligence put it plainly: "Pittsburgh Pirates at Atlanta Braves: Sportsbooks, Kalshi, and Polymarket are all aligned within 2pp — efficient market."

The largest pairwise gap across the three venues (Polymarket at 56.5% vs. Pinnacle at 58.8%) is 2.3 percentage points. That is a convergence story, not a divergence one. The markets largely agree on who wins.

You can pull all three probability streams programmatically through the Data API if you want to run your own cross-market alignment checks.

What the line movement says

This time, the market is not fighting the signal. It is reinforcing it.

The h2h line moved 84 pips up ahead of game time. Up means toward Atlanta. The favorite got a touch more expensive into first pitch, the same direction the XGBoost and composite scores already pointed.

This was not a Pinnacle-led move. In a classic sharp-money steam scenario, Pinnacle opens the door and the rest of the books follow. That did not happen on the h2h here: pinnacle_moved_first is false and the lead time is null. So the drift toward Atlanta is real but not stamped as sharp origination. No persistent reverse line movement is flagged either.

The run line moved the same way: 54 pips up, toward Atlanta's -1.5 side. Both sides of the Atlanta number drifted together.

The total is where the sharpest book did lead. The Over moved 455 pips up, and on totals pinnacle_moved_first is true with a 120-minute lead. Pinnacle pushed the Over first and the rest of the market followed it up toward the current 8.5 / 9.0 split. Given that both G1 and G2 hit exactly nine runs, there is some recency logic behind Over action, though markets tend to correct for that kind of pattern quickly.

What the whales are doing, and why this reading requires care

GhostLeg's whale consensus feed tracks Polymarket smart-money wallets and surfaces a directional read when wallet count and volume clear its thresholds. On this game it did fire: 80 wallets and $3,427 in tracked volume, with the consensus tagged high confidence.

But read the dollar figure before the label. $3,427 across 80 wallets is a thin Polymarket pool for a regular-season MLB game. The NBA playoff games that draw 500 to 800 wallets and six figures in volume (the Spurs-Thunder series, for instance) are a fundamentally different data environment. The wallet count here is real, the dollars behind it are small, and the underlying Polymarket perspective on this market is unresolved in our data.

So treat the whale layer as low-weight confirmation at best, not a standalone directional call. On a game like this, the h2h, run-line, and totals market signals carry far more weight than prediction-market wallet activity at this depth.

Series context and head-to-head history

Atlanta leads this weekend series 2-0. Both wins were 6-3. The Pirates need Sunday's game to salvage the finale and avoid being swept out of Atlanta.

Over the past 18 months (from 2025-05-09 through 2026-06-06, covering 8 games), the Atlanta Braves and Pittsburgh Pirates are exactly 4-4 against each other. But that 4-4 record includes Friday and Saturday's games from this series, both Atlanta wins. Before this weekend, the matchup was 4-2 in Pittsburgh's favor. Atlanta has won the last two in a row to flip the ledger to even.

That trend reversal is recent enough to matter. The Braves have found something in this particular matchup at this particular moment in the schedule.

Last-5 form and a notable tension in the numbers

Recent form adds a wrinkle worth sitting with. Over the last five games, both teams are 3-2.

Atlanta's five-game run differential: +0.2 per game (4.6 runs scored, 4.4 runs allowed). The Braves have been close to neutral on run production, winning tight games.

Pittsburgh's five-game run differential: +1.8 per game (7.2 runs scored, 5.4 runs allowed). The Pirates have been the better offensive team by run differential over this window, and it is not especially close.

Yet Pittsburgh has lost both games in this series. The script says Atlanta is winning efficiently while Pittsburgh's run production is arriving in the wrong moments, or in blowout wins that pad the differential without changing the series picture. That kind of pattern tends not to be durable. Whether it corrects on Sunday, in front of a home crowd that wants the sweep, is one of the more genuine questions in this matchup.

The full signal picture across all three markets

The h2h composite of 58 is the strongest reading of the three, sitting exactly where the XGBoost edge score landed. Spreads and totals show moderate signals. All three markets tilt the same general direction: Braves to win, game likely to produce runs.

What stands out about this game is the absence of friction. The XGBoost layer, the composite, the h2h and run-line movement, and the three prediction-market prices all point at Atlanta, and they cluster inside a couple of points of each other. The one genuine counterweight is Pittsburgh's last-five run differential, which has been the better of the two by a clear margin even as the Pirates dropped both games of this series. Whether that production starts converting into wins on Sunday is the live question.

Track how this line continues to move before first pitch in the MLB intel dashboard. The Data API surfaces the same cross-market streams for anyone running their own models; the licensing details and endpoint specs are at ghostleg.ai/data-api.

For entertainment purposes only. Past performance does not indicate future results.