Daily Dev Update: Smarter Bots, Clearer Wins - August 29, 2025

Period: August 29, 2025 Total Commits: 9

Hey, here’s what we’ve been working on today. The headline: bots got a real brain upgrade, tower placement got safer and smarter, and we simplified how matches end to keep PvP moving. A few rough edges still, but it’s getting there.

What’s New

• New multi-threaded bot system

  • Bots can now join lobbies, pick initial units, place towers adjacent to castles or existing towers, and assign minions to valid spawns.
  • There’s a small REST layer so we can spawn/stop bots and check status while we iterate.
  • The goal is better practice matches and more consistent lobbies while we tune PvP balance.

• Simpler victory condition

  • Win when the final castle falls. You no longer need to wipe every intermediate castle.
  • This shifts gameplay toward decisive pushes and clearer win/loss moments. Less grind, more clutch defenses.

• Stage-aware tower selection UI

  • Towers are now tagged with a gameplay stage (early/mid/late) and show up with color-coded indicators.
  • Helps you pick the right tools for where the match is at. Bots use this too—only early towers at level 1, then scale up appropriately.
  • [screenshot: tower selection screen with stage indicators at ba9cd73]

• Smarter frontline focus

  • Bots prioritize defending alive castles on the left side first (front-line around x:15), building outward from existing towers to create real choke points.
  • This keeps early skirmishes meaningful and reduces wasted gold on dead castles.
  • [screenshot: towers clustering around front-line castles at daf63fc]

What Got Fixed

• Tower placement is now path-safe and overlap-free

  • We validate the path so towers can’t block minions.
  • No more overlapping towers; placement checks the authoritative grid before committing.
  • Internal position tracking is more robust; we track both the registry and the grid to stay in sync.
  • [screenshot: updated tower placement grid showing no overlaps at 00c397a]

• Castle and anchor selection

  • Bots consider all alive castles (not just the bottom one) and use nearby living defenses as anchors for clustering.

• Gold accuracy and spend safety

  • Bots now initialize with the correct starting gold and stay in sync via SimCore events.
  • No more negative gold. We fixed pending spend handling so failed timeouts can’t drive the balance below zero.

• Mid-game joins don’t stall unit selection

  • Bots that join a match in progress now select their starting towers and minions correctly, even if the first wave already started.
  • Better logging and a clean retry path if transactions fail.

• Minion targeting clarity

  • Minions now prioritize the shortest path front-to-back, aligning with the simplified victory condition.
  • [screenshot: match end banner when the final castle falls at a58f03d]

What’s Still Broken

• Tower clustering can still pick a suboptimal tile in rare edge cases when multiple anchors compete at once. We’re tuning the tie-breakers.
• Under heavy packet loss, bots sometimes hesitate after a failed transaction before retrying. The safeguard is working, it’s just a bit conservative right now.
• The stage sorting UI for towers does what it says, but we’re still polishing the order for a few outlier tower types. Color coding is solid; the sort logic needs a pass.

If you see odd pathing around tight choke points, send us a clip. Those reports help.

Under the Hood

• Bot architecture

  • Worker threads per bot, with dynamic CPU core detection so we don’t swamp your machine.
  • Full game state sync via blockchain events, piped through a chain reader into SimCore for deterministic decisions.
  • Rest endpoints to spawn/stop/status are in place for development workflows.

• Economic consistency

  • We standardized on reading gold states from SimCore and added strict pending spend rules (always cleared on completion/failure, never dips below zero).
  • Player join events are processed during replay, so bots entering mid-match get the right baseline.

• Match lifecycle

  • The bot stack listens for the match end, shuts down decision loops, cleans up workers, and frees resources. Less ghost logic after GG.

• Placement rules

  • Valid placement is now adjacent-only, snapped to the grid, and checked against an authoritative occupancy map. That closed several desync and overwrite cases.

Why this matters for PvP

• Faster conclusions with the final-castle win condition means more matches, less stalemate.
• Smarter tower staging reduces early-game noise and creates clearer power spikes.
• Reliable bot opponents make it easier to test strategies and tower synergies without waiting for a full lobby.

What’s Next

• More aggressive anti-blocking heuristics so creative mazes don’t accidentally jam lanes.
• A pass on bot risk tolerance: earlier pressuring with minions when front-line defenses are light.
• UI tweaks for tower lists: tighter sorting, better tooltips, and clearer upgrade breakpoints.

Thanks for the reports and screenshots—especially on wonky placements. Keep them coming. We’ll keep shipping small, concrete improvements day by day.