The 100:1 Time Odds Massacre

[goni-K26]

The 100:1 Time Odds Massacre | Is the "Magic" of Houdini 6 Officially Dead?

Tournament Setup:

• Engine 1: Houdini 6.03 x64 (BMI2) | TC: 15m + 10s (Rapid) | Hash: 256 MB
• Engine 2: Stockfish 18 x64 (AVX2) | TC: 9s + 0.1s (Ultra-Bullet) | Hash: 16 MB
• Hardware: Intel(R) Core(TM) i5-4570 @ 3.20GHz | 1 Thread
• Ratio: Exactly 100:1 Time Odds

The Final Score (100 Games):

Code: Select all

Stockfish 18:  65.5  (41 Wins, 49 Draws, 10 Losses)
Houdini 6.03:  34.5  (10 Wins, 49 Draws, 41 Losses)

The Analysis:

I decided to conduct a brutal experiment to see if a massive 100-to-1 time advantage
could save a classical legend from the modern NNUE slaughterhouse.
Most people say time is the great equalizer in chess. Well, after 100 games, I can tell you: Time is dead. Logic is dead.
There is only Stockfish.

1. 9 Seconds vs 15 Minutes:
Stockfish 18 had exactly 9 seconds for the entire game. Houdini 6.03 had 15 minutes. Let that sink in.
Stockfish was playing at the speed of light, with a tiny 16 MB Hash (basically thinking with its eyes closed), and it still treated a former
World Champion like a club player.

2. The Evaluation Gap:
Even with 100x more thinking time, Houdini’s hand-coded evaluation was blind to the deep positional traps of the NNUE era.
It’s not just a strength gap; it’s a biological evolution. Houdini looks like a calculator, while SF 18 looks like a God.

3. A 13.5% Disaster:
In my wider 'Elite-25' tournament, Houdini finished dead last with a pathetic 13.5% score.
Even with 5x time odds against the whole field, it couldn't survive. But this 100:1 match against SF 18 is the final nail in the coffin.

The Provocation:

Is Robert Houdart’s masterpiece now officially a "legacy engine" that belongs in a museum? Or is Stockfish 18 so efficiently optimized that time odds have become a joke?

Warning: If you are a fan of classical 'Human-like' evaluation,
the PGNs of this match will break your heart. Houdini wasn't out-calculated; it was out-classed in every single phase of the game.

What do you think? Is there any amount of time that can make Houdini competitive again, or has the "Magic" finally run out?

[mar]

what are you talking about? Houdini 6 was an illegal Stockfish derivative (H6 source was leaked some time ago)

[goni-K26]

what are you talking about? Houdini 6 was an illegal Stockfish derivative (H6 source was leaked some time ago)

Listen Martin, I’m well aware of the history behind Houdini 6 and the source code controversy. That’s old news for anyone who follows the scene. However, you're missing the point of my experiment.

The focus here isn't on the "legality" or the "origin" of the code, but on the sheer technological leap we've seen since that era. Back then, whether the code was "borrowed" or not, Robert Houdart managed to optimize it in a way that gave it a significant Elo boost and a very specific positional "feel" that dominated the field for years.

The 100:1 Time Massacre was a test of evolution, not a legal audit. I gave Houdini 15 minutes against 9 seconds of modern Stockfish 18 on a single thread (Intel i5-4570). The fact that a 100x time advantage can’t bridge the gap between a classical legend and a modern NNUE engine is what’s fascinating here.

I’m a fan of the performance and the history, not the license agreement. It’s about the end of an era. The "Magic" is officially dead not because of the code’s history, but because modern engines have reached a level where time odds simply don't matter anymore.