Matrix Games

What’s New in Beta 3

Beta 3 is finally ready. This update primarily focuses on optimization and general improvements based on your feedback.

Originally, the plan was to emphasize the “nuclear” aspect of the game more heavily. However, smoothing out the game “stuttering” caused by the significantly increased CPU demands of the updated AI required far more effort than anticipated. While working on the planned optimization steps - and thanks to the save files many of you sent - I discovered additional opportunities to further improve performance, especially in the late game. As a result, I decided to prioritize performance improvements for this update.

I genuinely hope you’ll feel the difference.

That said, because the changes to the underlying execution engine are substantial, there is unfortunately a higher risk of occasional crashes. If you encounter one, please follow these steps:

1. Go to Documents → My Games → ICBM-Escalation

2. Locate the file Fatal.txt and any crash dump files (something like crashinfo_ver. 1.1.2 beta 3_2026.02.16__13.20.51.dmp)

3. Compress them into an archive (.zip, .rar  -  any format works)

4. Post the archive along with a description of the issue on the game’s Discord server or email it to vufnarovsky at gmail 

If you are able to reliably reproduce a crash, that would be enormously helpful.

AI Improvements

Another major focus of Beta 3 was AI aggressiveness - while also improving its ability to understand the overall strategic situation.

In this version, the AI is once again “more fun”, behaving closer to Beta 1 in how it pushes advantages. When an invasion succeeds, it now continues the offensive instead of stopping after capturing a single region.

Beta 2 introduced an issue that made the AI overly cautious when initiating invasions. It treated all non-friendly assets as hostile - not just the factions it intended to attack. This caused the AI to delay invasions until it was strong enough to handle not just one opponent, but all potential enemies combined. This clearly limited its effectiveness. That behavior has now been corrected.

The AI in Beta 3 is more flexible in choosing what to build. As example, it may now construct a full airbase instead of a tactical one if aircraft range would otherwise be insufficient to support an invasion. It also evaluates the “efficiency” of different units for specific roles, for example, sometimes preferring MRBM TELs over SRBMs, Cruiser over Destroyer, and vice versa, depending on the situation.

Strategic Awareness & Diplomacy

For the first time, the AI tracks the history of various events and incorporates that data into diplomatic decisions.

One practical improvement is that it should be less repetitive and less likely to spam the same diplomatic requests. More importantly, it now performs a linear regression analysis of military strength trends among factions. This allows it to assess whether it is gaining or losing momentum in a conflict.

Additionally, the AI now distinguishes between a faction’s conventional forces and its nuclear capabilities. It attempts to estimate the probability of full-scale nuclear war and adjusts its evaluation of overall faction power accordingly - depending on whether it expects the conflict to remain conventional or escalate.

Looking Ahead to Beta 4

With much of the foundational work now in place, Beta 4 will focus heavily on the nuclear dimension of the game.

My hope is that Beta 3 strikes the right balance between “too aggressive” and “too cautious,” setting the stage for Beta 4 to deepen the AI’s focus on mid-to-late-game missile defense and nuclear deterrence.

Changelog

Here below you can find the latest changelog:

1.1.2 beta 3

• Fixed the problem with too many Ballistic Missile Sites/SRBMs built-in,

• Fixed the problem with the AI being too cautions with invasions,

• Fixed the problem with the AI agreeing on ceasefire too quickly even when winning,

• Fixed the game "stuttering" when the AI builds logistics routes,

• Fixed the game "stuttering" when the AI analyses the overall map and adjusts its strategy,

• Fixed the game "stuttering" when the AI was deleting its dynamic plans,

• Fixed a problem with the AI being able to start war during peace time,

• Generally increased the game performance,

• "Priority research complete" journal event is now triggered only when the research queue becomes empt,

• The AI is now less prone to "spam" player with the same requests,

• Changes the way AI evaluates the faction's strength,

• The AI now tracks the war progress and takes decisions based on the overall success instead of the momentary force distribution