# Suppose 3450 AlphaZero Gives 100:1 Time Odds to 2835 Magnus. How would Magnus Do?

(I don't know if AlphaZero is really 3450 Elo.  It just seemed like a reasonable number.)

AlphaZero is back in the news.  Crushing StockFish.  Even at 10:1 Time Odds.

I know that if I had 2 hours to make all my moves against Magnus who only has 5 minutes for all his move, he'd still crush me.  After all, he can think on my time.

Now I'm wondering what if AlphaZero gave Magnus 100:1 time odds.  Let's say Magnus gets 120 minutes for all his moves with a 30 second increment, while AlphaZero gets only 1.2 minutes for all its moves with no increment.

Do you think Magnus can score some points?  How about 1 draw, at least?

What if it was StockFish?  How would Magnus do against StockFish at 100:1 Time Odds?  Score some points?  At least one draw maybe?

Thoughts?

Too scary/sterile for me. I like humans being humans, battling humans.

kamalakanta wrote:

Too scary/sterile for me. I like humans being humans, battling humans.

Understand.  However, let's remember that the great Garry Kasparov.took on Deep Blue back in the day on fairly even terms.

How would Magnus do if he were getting 100:1 time odds?

P.S.  My guess is that if it were a 6 game match, he'd be lucky to get a draw even at 100:1 time odds.

I am so happy that we cannot quantify the Infinite!

kamalakanta wrote:

I am so happy that we cannot quantify the Infinite!

I've taken an Upper Division math course in my university days, and there are Math Equations with various kinds of infinities.  It was a Set Theory Class, and it was hard.  I remember having to prove parts of Godel's Incompleteness Theorems, and it was pretty tricky.

I think it would depend on whether the machine can think on Magnus's time and how long it is given to make each move by whoever is running it.  1.2 minutes is 72 seconds and with no increment Magnus might be able to flag it.  If it spends 2 seconds per move Magnus only has to get to move 37 to win on time.  I really have no idea how computers ration their time.  Most programs I've seen take a set amount of time per move, but not per game.

Assuming the machine doesn't flag, it would probably win.  Magnus could beat me blindfolded playing one second on a five second delay and I could be playing correspondence TC .

SeniorPatzer wrote:
kamalakanta wrote:

I am so happy that we cannot quantify the Infinite!

I've taken an Upper Division math course in my university days, and there are Math Equations with various kinds of infinities.  It was a Set Theory Class, and it was hard.  I remember having to prove parts of Godel's Incompleteness Theorems, and it was pretty tricky.

Still, there are gaps in our ability to quantify the infinite. We don't know if aleph-one and the cardinality of the continuum are one and the same, or if the latter is larger.

SeniorPatzer wrote:
kamalakanta wrote:

I am so happy that we cannot quantify the Infinite!

I've taken an Upper Division math course in my university days, and there are Math Equations with various kinds of infinities.  It was a Set Theory Class, and it was hard.  I remember having to prove parts of Godel's Incompleteness Theorems, and it was pretty tricky.

Still, there are gaps in our ability to quantify the infinite. We don't know if aleph-one and the cardinality of the continuum are one and the same, or if the latter is larger.

Does Infinity exist?

knighttour2 wrote:

I think it would depend on whether the machine can think on Magnus's time and how long it is given to make each move by whoever is running it.  1.2 minutes is 72 seconds and with no increment Magnus might be able to flag it.  If it spends 2 seconds per move Magnus only has to get to move 37 to win on time.  I really have no idea how computers ration their time.  Most programs I've seen take a set amount of time per move, but not per game.

Assuming the machine doesn't flag, it would probably win.  Magnus could beat me blindfolded playing one second on a five second delay and I could be playing correspondence TC .

Lol, that would be just hilarious to watch Magnus trying to flag AlphaZero!

But yes, AlphaZero does get to "think" on Magnus's time.  And the clock would have to be set up in tenths of a second, and AlphaZero can thereby make some of its moves in tenths of a second in order to not flag.  After all, 72 seconds is just 72 moves if the smallest time unit expended is a second.  Therefore, the smallest time unit will be set to tenths of a second.

No computer can calculate God's infinite Love.

Please notice how now the creative element of chess has ceased to be important, and the competitive element is paramount....win/lose, etc.

No more Brilliancy Prizes....I don't know what to say. I cannot relate to all this.

rychessmaster1 wrote:

At 100:1 time odds, I would beat anyone

WITH THE 1

You get 500 minutes with no increment/delay against Magnus Carlsen with 5 minutes and no increment/delay.

Magnus would win.

AZ has no chance against the greatest ever since the dawn of time until the end of the cosmos.

rychessmaster1 wrote:
SeniorPatzer wrote:
rychessmaster1 wrote:

At 100:1 time odds, I would beat anyone

WITH THE 1

You get 500 minutes with no increment/delay against Magnus Carlsen with 5 minutes and no increment/delay.

Magnus would win.

Maybe, but i'm good

Against everyone but me maybe. you have a minute and i have 10, i win

play me 3/2 get on live

1) Stockfish 10  is still the best engine in the world
2) LC0 guys did manage to reverse engineer A0 successfully
3) LC0 and A0 roughly at the same strength
4) NN are not less resource hungry than Alpha Beta
5) Scalability is about the same in both methods
6) Google has unacceptable behaviour, hiding data, obfuscating opponents and hyping results

Alpha Zero won SF8 by +50 elo only.( AlphaZero vs Stockfish 8 in a new 1,000-game match, scoring +155 -6 =839) So Alpha Zero is roughly the same strength as SF 9.

SeniorPatzer wrote:

Does Infinity exist?

actual infinities can't exist but abstract ones certainly do ?

natarka wrote:
SeniorPatzer wrote:

Does Infinity exist?

actual infinities can't exist but abstract ones certainly do ?

What is your basis for saying that actual infinities can't exist?

SeniorPatzer wrote:

What is your basis for saying that actual infinities can't exist?

Tom Campbell (author of "My Big TOE") says so and the universe itself is arguably finite - give me a few beers however and I might change my mind about all of it ...

For Stockfish and AB engines, they lose 50-60 elo  , by every half time handicap, but A0 and Leela will lose much less elo than Stockfish ( because their strength is mainly based on NN decision, not the search)and A0 would probably lose about 25 elo in every half time handicap. So for 100 time odd( 2^7), A0 will lose rating between -350 to -175 and would have 3100 -3200 rating.

So, A0 would crush 2800 Magnus by 85%-90% scores.