Back

BGH Group Batch Coach Report v0.2

Date: 2026-06-02

Update: added Jun's broader replay aliases:

Corrected Sample

The BGH replay count remains 433.

After alias expansion, Jun-related samples increase from 57 games to 362 games.

PurE)Rabbit-SF is the largest main sample:

After removing ultra-short remake/drop games, Jun aliases have 327 meaningful-length samples.

Meaningful samples by race:

RaceWinsLossesUnknownAvg APM/EAPMFirst read
Protoss53554225 / 180closest to stable baseline
Zerg43604233 / 179active, but not converting enough
Terran336510229 / 184first repair target

Corrected First Read

The previous report only saw Dimsumsum[STA], which undercounted the real sample and overestimated Zerg.

After including PurE)Rabbit-SF / UncleTiger, the read changes:

  1. Jun is a long-term random/multi-race BGH player.
  2. APM/EAPM is high enough across all races; speed is not the first issue.
  3. Terran is the clearest weakness.
  4. Protoss is the best baseline race to stabilize first.
  5. Zerg has activity, but needs a conversion review.

Priority 1: Make Protoss The Baseline

Protoss is nearly even at 53W-55L-4U.

That makes it the best first coaching baseline:

Training rule:

In every Protoss BGH game, define your role first: front defense, side pressure, or late-game core.

Review questions:

  1. Was my spawn front-line or back-line?
  2. Did my first Pylon/Gateway/Forge/Cannon shape serve the team defense?
  3. Did my first Zealot/Dragoon movement protect an ally, pressure an enemy, or wander without purpose?
  4. Did I become a team core by midgame?

Priority 2: Terran Is The Repair Lane

Terran is the clearest weak sample: 33W-65L-10U.

The mechanical baseline is not low: around 229 / 184 APM/EAPM.

The likely issue is that Terran activity is not reliably converting into BGH team value:

Terran target:

Prove team value before 6-8 minutes.

Review standard:

Priority 3: Zerg Is A Conversion Problem

Zerg is 43W-60L-4U with around 233 / 179 APM/EAPM.

This is not slow Zerg. It is a conversion question:

Did the first Zerg activity actually change the team game?

Zerg review should check:

Training target:

In every Zerg game, identify whether the first unit wave changed the team position.

Product Direction

This corpus is enough to start a real BGH coach.

The coach can already produce:

The next product is not generic Brood War AI.

It is:

WeChat BGH group match report + coaching system

Core loop:

  1. Upload a replay batch.
  2. Detect group aliases.
  3. Produce player profiles.
  4. Identify strong/weak races.
  5. Select the top 3 review candidates.
  6. Give friend-readable coaching advice.

Immediate Next Deep Dives

Terran repair:

Protoss baseline:

Zerg conversion:

Coach Verdict

With aliases corrected, the conclusion is stronger:

Jun already has a long-term multi-race BGH corpus.

The first AI coach should:

  1. unify PurE)Rabbit-SF / Dimsumsum[STA] / UncleTiger identity
  2. use Protoss as the stable baseline
  3. treat Terran as the first repair lane
  4. investigate Zerg conversion separately

Human version:

You do not lack speed. You have played too much random-race BGH, and Terran team value is not consistently converting. Build the baseline with Protoss, repair Terran, then inspect why Zerg activity is not winning more.