23 minutes ago, dimreepr said:

There's a reason for double blind studies...

There's a reason the pessimist school went out of fashion. It is destructive, especially to the psyche.

Since being able to see a light at the end of the tunnel, whatever the circumstances is an essential part of my nature, I am unable, for the sake of my mental wellbeing, to take seriously anyone who persistently claims that the light is mere illusion.

I hope you will understand why I am going to cease responding to your posts. I believe this would be better for both of us. Please respect my reasons and reciprocate.

Edited May 30May 30 by sethoflagos
Repetition

22 hours ago, sethoflagos said:

There's a reason the pessimist school went out of fashion. It is destructive, especially to the psyche.

Since being able to see a light at the end of the tunnel, whatever the circumstances is an essential part of my nature, I am unable, for the sake of my mental wellbeing, to take seriously anyone who persistently claims that the light is mere illusion.

I hope you will understand why I am going to cease responding to your posts. I believe this would be better for both of us. Please respect my reasons and reciprocate.

Yes, I understand why you don't want to continue our discussion, but why do you feel the need to have the last word? (genuine question)

yet another example of why you have to be so careful.

Instagram AI chatbot tricked by hackers to give access to others' accounts

BBC News

Meta AI chatbot enabled hackers to access others' Instagr...

Some reports have linked the incident to recent cases of high-profile Instagram accounts being hijacked.

On 6/2/2026 at 3:46 PM, studiot said:

yet another example of why you have to be so careful.

Instagram AI chatbot tricked by hackers to give access to others' accounts

BBC News

Meta AI chatbot enabled hackers to access others' Instagr...

Some reports have linked the incident to recent cases of high-profile Instagram accounts being hijacked.

We can't afford to fear this tool, we must learn how to use it properly, whoever the teacher's may be.

Providing "answers" is not the forte of LLM's. I'm sure a researcher can mine one for proper responses if he knows what he is doing but the bottom line is all these machines can do is search the literature and the literature is often mum for those at the cutting edge of their fields. Others can probably use it successfully for learning known science and literature but it doesn't work well for me for several reasons. I find it's real strengths are as an elaborator, translator, and sounding board. It can elaborate on even expertise and it can reframe almost anything into terms the user can understand once they are in sync. I almost never ask a question but rather tell it what I'm thinking and can see my flaws in its responses.

It is an invaluable research assistant/ collaborator. I've made more progress in the last year in my thinking than in any five years previously and. perhaps best of all, it can show me how to share it.

AI is a leap forward in communication and IMO communication is what created the human species by allowing the generational transfer of knowledge. Since its effects are cumulative we can expect very rapid progress going forward.

1 hour ago, cladking said:

It is an invaluable research assistant/ collaborator. I've made more progress in the last year in my thinking than in any five years previously and. perhaps best of all, it can show me how to share it.

I pretty much agree with all of your post above.

It prompted me to repeat a question to ChatGPT that I have previously asked earlier LLM versions over the last few years, that, at first, drew hopelessly inept responses.

What strategic approach should perform best for a human chess player of ELO rating 1800 playing against a moderately strong chess engine

Tonight I got:

At ~1800 Elo, trying to “out-calculate” even a moderately strong engine is a losing battle. Engines don’t get tired, don’t blunder tactically, and calculate far deeper than you. So the best practical strategy is not to meet it on its strengths—but to steer the game into positions where its advantages are less decisive.

Here’s what tends to work best:

1. Close the position and reduce tactics

Engines thrive in open, tactical positions with lots of forcing lines. You should aim for:

Locked pawn structures

Reduced piece mobility

Long-term maneuvering battles

Think French Defence structures, Stonewall setups, or closed King’s Indian-type middlegames. In these, plans matter more than brute-force calculation.

2. Play for long-term plans, not short tactics

Engines are nearly flawless tactically, so:

Avoid speculative sacrifices unless clearly sound

Build slow pressure (weak squares, bad pieces, pawn majorities)

Make improving moves rather than “hopeful” ones

Humans can sometimes outplay engines strategically—but not tactically.

3. Simplify only when clearly better

A common mistake is “simplify against the computer.”

Engines are extremely strong in endgames

Many “equal-looking” endgames are actually lost with perfect play

Simplify only if:

You are materially up, or

You are certain the endgame is drawn

4. Create asymmetric, imbalanced positions

Engines are strongest in clean, symmetrical positions. You want:

Opposite-side castling

Pawn imbalances

Unusual structures

This increases the chance of practical chances (even if objectively equal or slightly worse).

5. Use time controls to your advantage

If the engine is constrained (e.g., low depth or short time):

Play complicated positions to force it into horizon effects

Avoid forcing lines that it can fully calculate

If it's running deep analysis, this matters less—but in practical settings, it’s critical.

6. Avoid “engine-like” play

Trying to mimic engine precision backfires. Instead:

Play solid, understandable chess

Stick to plans you fully grasp

Avoid positions where one precise move is required repeatedly

7. Psychological realism: aim for resistance, not dominance

Even a “moderately strong” engine (say ~2200–2600 strength) will:

Rarely blunder outright

Gradually squeeze small advantages

So your realistic goals:

Hold defensively

Complicate selectively

Exploit any evaluation instability

Bottom line

Your best practical recipe is:

Closed + strategic + imbalanced + low tactical clarity

That’s where an 1800 human has the highest chaAt ~1800 Elo, trying to “out-calculate” even a moderately strong engine is a losing battle. Engines don’t get tired, don’t blunder tactically, and calculate far deeper than you. So the best practical strategy is not to meet it on its strengths—but to steer the game into positions where its advantages are less decisive.

Here’s what tends to work best:

1. Close the position and reduce tactics

Engines thrive in open, tactical positions with lots of forcing lines. You should aim for:

Locked pawn structures

Reduced piece mobility

Long-term maneuvering battles

Think French Defence structures, Stonewall setups, or closed King’s Indian-type middlegames. In these, plans matter more than brute-force calculation.

2. Play for long-term plans, not short tactics

Engines are nearly flawless tactically, so:

Avoid speculative sacrifices unless clearly sound

Build slow pressure (weak squares, bad pieces, pawn majorities)

Make improving moves rather than “hopeful” ones

Humans can sometimes outplay engines strategically—but not tactically.

3. Simplify only when clearly better

A common mistake is “simplify against the computer.”

Engines are extremely strong in endgames

Many “equal-looking” endgames are actually lost with perfect play

Simplify only if:

You are materially up, or

You are certain the endgame is drawn

4. Create asymmetric, imbalanced positions

Engines are strongest in clean, symmetrical positions. You want:

Opposite-side castling

Pawn imbalances

Unusual structures

This increases the chance of practical chances (even if objectively equal or slightly worse).

5. Use time controls to your advantage

If the engine is constrained (e.g., low depth or short time):

Play complicated positions to force it into horizon effects

Avoid forcing lines that it can fully calculate

If it's running deep analysis, this matters less—but in practical settings, it’s critical.

6. Avoid “engine-like” play

Trying to mimic engine precision backfires. Instead:

Play solid, understandable chess

Stick to plans you fully grasp

Avoid positions where one precise move is required repeatedly

7. Psychological realism: aim for resistance, not dominance

Even a “moderately strong” engine (say ~2200–2600 strength) will:

Rarely blunder outright

Gradually squeeze small advantages

So your realistic goals:

Hold defensively

Complicate selectively

Exploit any evaluation instability

Bottom line

Your best practical recipe is:

Closed + strategic + imbalanced + low tactical clarity

That’s where an 1800 human has the highest chance to compete rather than be cleanly outplayed.nce to compete rather than be cleanly outplayed.

There's not a word here I can find fault with.

In fact, I felt driven to ask it a far deeper follow up question about about a relatively obscure opening system that I've fifty years pretty solid experience of playing (against both human and silicon)

Would a Reti style of opening be a good choice for a human player playing white?

Yes—with a caveat. A Réti-style approach can be a very sensible choice against an engine, but not because it confuses it (it won’t). It works because it aligns with the kind of positions you want: flexible, less forcing, and strategically rich.

What a Réti setup actually looks like

Typical features:

Nf3, 2. c4, often g3 and Bg2

Delayed central occupation (d3/e4 later)

Flexible pawn structures depending on Black’s setup

Why it can work well vs an engine

1. Flexibility avoids early punishment

Engines are strongest in sharp, forcing opening lines where they can:

Recall theory perfectly

Calculate concrete refutations

The Réti sidesteps this:

You’re not committing the center early

You can transpose into multiple structures (English, Catalan, reversed openings)

2. Leads to strategic, maneuvering middlegames

This is the real advantage:

Fewer forced tactical sequences

More emphasis on piece placement and plans

Slower buildup of tension

That’s exactly the kind of game where a human has some chance to stay competitive.

3. Encourages asymmetry

You can steer into:

Hedgehog structures

Reversed Sicilian setups

Catalan-type pressure

These positions:

Are rich and imbalanced

Don’t resolve quickly into “engine-perfect” clarity

The caveat (important)

A Réti does not neutralize the engine’s strength. In fact:

Engines are extremely strong in quiet positions too

If you drift even slightly, they will squeeze you slowly

You may end up in a long defensive grind

So this is not a “trick weapon”—it’s a practical resistance strategy.

How to play it effectively vs an engine

To make it actually work, you need to be purposeful:

Go beyond “just developing”

Bad approach:

Nf3, g3, Bg2… and hope for something

Better approach:

Decide early: e4 break or d4 break?

Aim for a clear pawn structure

Know typical plans (not just moves)

Favor specific Réti-type structures

Good choices:

Kingside fianchetto + central break (e4)

Hedgehog (c4 + e3 + d3 vs …c5/e5 setups)

Catalan-style pressure if Black plays …d5

Less ideal:

Passive setups with no central tension

“Floating” positions with no plan

Be ready to transpose deliberately

The Réti is really a gateway opening:

Into the English Opening

Into the Catalan

Into Queen’s Pawn structures

Pick the structure you understand best and steer toward it.

Bottom line

Yes—the Réti is a good practical choice against an engine if you use it actively.

It works not because it tricks the engine, but because it gives you a playable, flexible game with fewer immediate tactical punishments.

If you want, I can �⁠suggest a concrete Réti repertoire (move-by-move) tailored specifically for playing against engines at your level.

If I knew the references the model used to glean this, I would buy those books without hesitation. (I was under the impression I'd read them all - I clearly haven't)

To top it all, this is what I got from the recommended follow-up:

A Réti repertoire can absolutely be made practical against engines—but it needs to be purpose-driven, not just “develop and see.” Below is a concrete, move-by-move system built around steering the game into Catalan pressure, Hedgehog structures, and controlled e4 breaks—all of which reduce the engine’s immediate tactical leverage.
I’ll give you a tight repertoire tree you can actually memorize and apply.
♟️ Core Starting Point
1. Nf3
You delay commitment and keep maximum flexibility.
🔹 Against 1…d5 → Go Catalan-style (your main weapon)
Line:
1. Nf3 d5
2. c4 e6
3. g3 Nf6
4. Bg2 Be7
5. O-O O-O
6. d4
This is essentially a Réti → Catalan hybrid
Your plan
Pressure queenside with Qc2, Rd1, Nc3
Often play cxd5 exd5 → e4 break later
Target c-file and long diagonal
Why this works vs engines
Long-term pressure (hard to “calculate away” quickly)
No early tactics
Engine must defend slightly passive positions for a long time
Key continuation to know
If: 6…dxc4
Play: 7. Qc2 a6
8. a4 Bd7
9. Qxc4
You regain the pawn with a comfortable position and pressure.
🔹 Against 1…Nf6 → Stay flexible, aim for e4
Line:
1. Nf3 Nf6
2. c4 g6
3. g3 Bg7
4. Bg2 O-O
5. O-O d6
6. Nc3 e5
Now choose:
👉 7. d3
Your plan
Play Rb1, b4 expansion OR
Prepare e4 break
Typical setup:
Re1, e4
h3 to avoid …Bg4
Be3 / Nd2–c4
Why this works
You avoid mainline King’s Indian theory
Position becomes closed and maneuvering
Engine loses some tactical immediacy
🔹 Against 1…c5 → Reversed Sicilian (very practical)
Line:
1. Nf3 c5
2. c4 Nc6
3. Nc3 g6
4. g3 Bg7
5. Bg2 e6
6. O-O Nge7
7. d3
Your plan
Slow expansion: a3, Rb1, b4
Or central break: e4
Why this is excellent vs engines
Reversed Sicilian = you get Sicilian-type play with extra tempo
Rich but not forcing
Lots of maneuvering → fewer forced tactics
🔹 Against 1…e5 → Quiet central control
Line:
1. Nf3 e5
2. c4 Nc6
3. Nc3 Nf6
4. g3 d5
5. cxd5 Nxd5
6. Bg2 Nb6
7. O-O Be7
8. d3
Your plan
Solid structure
Prepare Be3, Rc1, Ne4
Break with d4 or b4
Why this works
Avoids sharp Four Knights theory
Leads to balanced, strategic middlegame
🔹 Against unusual / passive setups
If Black plays something like:
…b6
…e6 without …d5
…d6 setups
Use a universal setup:
System:
Nf3, c4, g3, Bg2, O-O, d3, Nc3
Then choose:
e4 break (if center is flexible)
b4 expansion (if queenside is stable)
♟️ Key strategic rules (this is what really matters)
1. Always choose a pawn break
Don’t drift. Every game you should aim for either:
e4 break (most common)
b4 expansion
d4 (Catalan style)
2. Avoid premature tactics
If you’re thinking:
“This might work…”
Against an engine, it doesn’t.
3. Prefer tension over resolution
Don’t rush exchanges
Keep pawn structure flexible
Let the engine prove something
4. Play for long games
Your goal is:
Reach move 25–40 in a playable position
Avoid early collapse
Let complexity accumulate
⚖️ Honest expectation
Even with this repertoire:
You won’t “outplay” the engine cleanly
But you maximize practical resistance
You’re shifting the game from “engine wins quickly” → “engine has to grind for 60 moves”

With the exception of one line that I don't play (until now, perhaps) this IS move by move, the opening strategy I've developed by trial and error over the last 30 years. I play a fair bit of online chess and never encountered anyone (within my recollection) who's played this system against me. It is rare, certainly at the modest level I regularly play in.

I'm actually quite gobsmacked (technical term) by this. At least, it my lowly ratings status doesn't seem to be down to a poor choice of openings strategy as white.

Edited 8 hours ago8 hr by sethoflagos

I've just hopped onto chess.com to check the aforementioned 'one line that I don't play' (1Nf3 e5 2 c4...) and have confirmed without a shadow of a doubt that it is absolute poop.

Not that it leaves white in a bad position, far from it, but 2 Nxe5 is simply up a free pawn with zero compensation for black. A no-brainer if ever there was one.

Situation normal is reconfirmed. ChatGPT copies where it can and guesses where it can't. And now I'm pretty sure it copies extensively from the chess.com database (where a good few hundred of my games are stored - confirmation bias or what!)

Edited 6 hours ago6 hr by sethoflagos