New iOS app shows real-time EVs as you play

[tomker]

1. I'm glad that we're still on good terms but I admit that I'm having trouble interpreting your tone. For example, you referring to my use of DigitalOcean as a "claim," implying that it may not be true, i.e., that I might be lying. Why would you make such an unnecessary implication? And telling me that I was only using 1 core of a 16 core computer, contradicting my claim that I was using 16, when you really have no evidence of what I have or haven't done?

3. The reason my simulations are limited to a relatively small number of hands is because that's what's computationally feasible on a cell phone processor if you want immediate results.

4. I disagree. I have two bodies of code that generate EVs, one does it exactly by enumerating and summing probabilities and the other does it via simulation. My app uses one or the other depending on what the situation calls for, but I also have diagnostic/test code that runs both and compares the results to make sure there aren't bugs in either. The test code does simulations of 100k hands and reports an error condition if the algorithms differ by more than 0.01. These errors are exceedingly rare, which proves to me that Monte Carlo simulation with 100k hands is actually quite accurate.

Given that, I would respectfully suggest that if you aren't seeing the same convergence rate with your Monte Carlo algorithm, you might have some bugs in your code. I hope you don't take that as an unjustified insult--I invite you to download my app and check some of its EVs yourself if you don't believe me.

[tomker]

Incidentally, there is no such processor. And if there were, it would not run at 16 times the speed.

You seem to be coming at this discussion from the point of view that I don't know what I'm talking about or that I'm not very bright, and you don't seem to be trying too hard to give me the benefit of the doubt.

Indeed, I never said I rented a machine with one processor. I said I rented one with 16 cores. Who knows how many processors it had? Maybe 2. Maybe 4. I don't really care.

And since my code is a recursive enumeration algorithm that uses minimal stack, all of the computation should easily fit in the L1 cache of a processor core, or at least the L2 cache. So there's no contention for main memory or L3 cache, so it's entirely possible for my algorithm to scale perfectly and run 16 times faster on a 16 core machine.

[Norm]

For example, you referring to my use of DigitalOcean as a "claim," implying that it may not be true, i.e., that I might be lying.

I said no such thing and wish you would stop making paranoid-sounding assumptions. I was obviously referring to the claims made by DigitalOcean, and repeat that no such XEON processor exists.

The reason my simulations are limited to a relatively small number of hands is because that's what's computationally feasible on a cell phone processor if you want immediate results.

RIGHT. Which is why such calculations should NEVER be done on a mobile device.

I disagree. I have two bodies of code that generate EVs, one does it exactly by enumerating and summing probabilities and the other does it via simulation.

Again, if your sims give results that close with that few hands, it is absolute proof that the sims are incorrect. This is statistically impossible as the standard error is VASTLY higher.

Given that, I would respectfully suggest that if you aren't seeing the same convergence rate with your Monte Carlo algorithm, you might have some bugs in your code. I hope you don't take that as an unjustified insult

Of course I don't take such an astoundingly silly statement as an insult. NO statistician would agree with you.


Again, I am trying to be polite, despite your assumptions to the contrary. But, you simply don't know what you are talking about, whether it comes to statistics, processors, blackjack, Monte Carlo techniques, or evaluating forum responses. For once, stop trying to assume ulterior motives in what I am posting and actually read all the words as if they are meant to explain the errors that you have made.

[Norm]

You seem to be coming at this discussion from the point of view that I don't know what I'm talking about or that I'm not very bright, and you don't seem to be trying too hard to give me the benefit of the doubt.

Indeed, I never said I rented a machine with one processor. I said I rented one with 16 cores. Who knows how many processors it had? Maybe 2. Maybe 4. I don't really care.

And since my code is a recursive enumeration algorithm that uses minimal stack, all of the computation should easily fit in the L1 cache of a processor core, or at least the L2 cache. So there's no contention for main memory or L3 cache, so it's entirely possible for my algorithm to scale perfectly and run 16 times faster on a 16 core machine.

Again, there is no such processor. I have never heard of the company you are using -- sounds like they are exaggerating. There is an eight-core processor with 16 threads. Hyperthreading does not double the speed as there are still only eight cores. It increases speed, ONLY IF your code is multi-threaded (almost no applications are) and there are idle areas of the core that are not data dependent.

[tomker]

Again, there is no such processor. I have never heard of the company you are using -- sounds like they are exaggerating.

And again, what makes you think the computer only had one processor? Maybe it had two 8-core processors. Or four 4-core processors.

Ultimately it doesn't matter. You seem to think that I bought into some hyped-up marketing claims, ran my code on a no-name service, and called it good. In fact, my code constantly printed out performance statistics about how quickly it was running and I compared those statistics to what I get with my desktop computer, which has Ivy Bridge cores. The DigitalOcean computer was able to do the workload exactly as fast as 16 2.3GHz Ivy Bridge cores. I saw the numbers with my own eyes, so nothing you say on this thread will somehow convince me that my own code ran slower than it did in reality.

But whatever, I was simply making a friendly suggestion about something you might consider doing in the future, since I had a great experience with the service myself. But if all you're going to do is imply that I'm lying or that I don't know what I'm doing, then I rescind my suggestion and we don't have to talk about it again.

[tomker]

... Again, if your sims give results that close with that few hands, it is absolute proof that the sims are incorrect. This is statistically impossible as the standard error is VASTLY higher. ...

Like I said, you can say they're incorrect all you want. It would be absurdly easy for you to download the app, use it to calculate some EVs, and post the (supposedly incorrect) results for all to see.

You will need to buy the "Ultimate version" upgrade to use the Hand Analyzer feature. That costs $6. If you send me an email with your iTunes receipt proving that you paid for the upgrade, I will be happy to reimburse you for it. So this will only take a couple minutes of your time and zero of your dollars.

[Norm]

I probably should have deleted the original post as clear spam in the first place. But, I don't like doing that. I have moved the thread to the Disadvantage Forum. Discussion can continue here. Of course, everyone here knows that you cannot sim blackjack results in a few hundred thousand hands. The standard error would be so high that you wouldn't even know if a game was plus or minus EV. CVData can't even run that small a number. It can run 20 million per second (without cheating with subsets and an imaginary processor).

[tomker]

I probably should have deleted the original post as clear spam in the first place. But, I don't like doing that. I have moved the thread to the Disadvantage Forum. Discussion can continue here. Of course, everyone here knows that you cannot sim blackjack results in a few hundred thousand hands. The standard error would be so high that you wouldn't even know if a game was plus or minus EV. CVData can't even run that small a number. It can run 20 million per second (without cheating with subsets and an imaginary processor).

Now that you've called me a cheater and a liar, let's see you put your money where your mouth is.

I bet you $50 that Blackjack X calculates correct EVs.

Just name a set of rules, the cards of a hand, the dealer's up-card, and the high-low count that you want to see simulated. I will send you a screen shot of the EVs that Blackjack X calculates. If any of them are off by more than 0.02, I will pay you $50. If they are all within that margin, you pay me $50.

I am at a complete disadvantage here because you can download my app and run through as many hands as you need to to find one that it gets wrong. But that's my level of confidence that my EVs are correct.

Edit: I will also mention that I clearly can't cheat at this bet since anyone is free to download my app and confirm the EVs on their own iPhone/iPad.

[Norm]

In case anyone is interested, I looked into DigitalOcean. ALL hardware uses six-core processors. When you ask for a 16-core processor, this is actually 16 virtual cores using KVM. KVM allows you to allocate two virtual cores per physical core on a hyperthreaded PC. In this case, you are NOT actually getting the full use of 16 cores as threads in hyperthreaded cores interfere with one-another. Further, you share the physical cores with other users. RAM and disk are segmented. You get full use of the amount of storage requested; although you may share access to the ‘disks’. Cores are NOT segmented. Cores are over-allocated and shared with other users via time-slicing. Over-allocating is common is such services. KVM will allow 1:1, 2:1, 10:1, whatever ratio of virtual cores to physical cores desired by the service. It appears that DigitalOcean has thus far refused to answer questions about the logical core to physical core ratio despite requests for this info going back to 2012.

Of course none of this matters in the least unless your code is multi-threaded or you run lots of instances at once.

[Dieter]

Of course none of this matters in the least unless your code is multi-threaded or you run lots of instances at once.

Multi-instance - like MPI or PVM - would probably suit "cloud" better. (HTCondor is probably overkill.)

I normally understand "cloud" to mean "dynamically provisioned virtual machines", not unlike Amazon EC2.

I do not normally understand "cloud" to mean VPS, like DigitalOcean seems to be offering.

Amazing Hardware

All cloud servers are built on powerful Hex Core machines with dedicated ECC Ram and RAID SSD storage.

In full fairness, sometimes hex means 6 (like it should), and sometimes it's an abbreviation for hexadecimal (16). This time, it's 6, not 16. This means exactly nothing, since it's a VPS, not dedicated hardware - you are splitting the hardware's capacity among multiple VM's with varying processor demand.

[Norm]

I believe it is like EC2.

This site was originally on a VPS. Performance was too erratic and I moved to a dedicated server.

[tomker]

In case anyone is interested, I looked into DigitalOcean. ALL hardware uses six-core processors. ...

Okay. But I'll point out that nothing you "found" contradicts anything I said. Intel makes a bunch of 6-core Xeons. And I never said the machine I rented had ONE processor with 16 cores, and I never actually said that it wasn't a virtual machine with virtual cores. I do know (according to my logs) that running my job on DigitalOcean was ~3.5x faster than on my home computer which has a quad-core Ivy Bridge i7, which is commensurate with the difference in clock speed that I mentioned earlier.

But ultimately Norm, please enlighten us with your point. The reason I mentioned the "cloud" or DigitalOcean in particular is as a (hopefully) helpful suggestion. Last year I would have run compute-heavy tasks on my own computer but this year I found a service that saved me a lot of time for a few dollars and I wanted to share that with you. Instead of a simple "thank you" I seem to have inadvertently gotten into a flame war with you calling me a liar... can you please explain that?

Of course none of this matters in the least unless your code is multi-threaded or you run lots of instances at once.

Right...? Who would rent a multicore machine to run a single core workload? Are you saying that your code isn't multithreaded? I would be happy to give you some pointers about that if you wanted.

Also, can I assume that because you didn't take me up on my bet, that's a tacit agreement on your part that my EVs are correct?

[Three]

Are you saying that your code isn't multithreaded? I would be happy to give you some pointers about that if you wanted.

Explore the help Norm has given people on this site to get his software to run at its fastest on their hardware. His software is multithread and he explained how many threads you should use for your machine to get optimal speed out of it. Everyone here knows just how elite Norm's expertise is in the industry as does anyone that has a history in blackjack software or systems. You might want to humble yourself a little and maybe you can learn something that you will find useful. Norm hasn't attacked you but walked a line between vetting you with his expertise for all in the site and giving you useful suggestions.

Norm always lets other software designers post their products on the site. Many fail the vetting process and end up banned for repeatedly posting false claims. If I were you I would do my best to represent yourself well and have constructive interaction with Norm. I think you have shown a desire to do that here but may be letting ego get in the way of your goal here. For all those that aren't experts in software development etc Norm will set the tone for how they perceive you. Like it or not that is the realty here. Use that knowledge to make the most of the site. There are many software developers that pop in from time to time and often put their heads together to find the best ways to do things. Besides being a good market for what you are peddling this site could be great for you in terms of R&D.

That is if you don't mess it up for yourself.