More from The Dotabuff Blog
70 Kommentare
Mandalord

    @Cunnilinguistician: It's deliberately designed that way. In normal RNG which mimics real random, each random event occurs INDEPENDENTLY. Which means, if that's the case, there is chance that Phantom assassin will not have a single crit in the entire game (very unlikely, probability is very close to 0, but possible). There is also chance that he will crit every single attack (very unlikely too, but also possible). This situation leaves the result to luck: In a combat, PA can crit many times and rekt enemy team, or no crit at all and meet her doom. They don't want it happens that way. They want luck to be manageable.

    So, the solution is, there is a counter which counts PA attacks and reset when she crit. The first attack of counter will have very low crit chance, with next attack have higher crit chance. The crit chance reach 100% at 17 consecutive no-crit attack (if i'm not mistaken). PA will get 15% crit in the long run, when number of attack reach infinity. Therefore, crit chance is manageable as you see in the article.

    You are perfectly true in thinking there is correlation, cause crit chance of each attack is not independent. Therefore it's deliberatively called pseudo-random, not real random or just rng or something.

    This mechanics stayed with the game since beginning of dota 1, and they recreated it in dota 2 also. The method of abuse in this article is not new at all, just old knowledge. You can search for many documents describe this in detail, not only this patch and not only dota 2. Here is one of them: http://dota2.gamepedia.com/Pseudo-random_distribution. Some people say we can look at the source code of dota 1 map (Warcraft 3 custom map) and find detail function of the pseudo rng. I have never do it myself, because I dont understand code.

    You can do experiment yourself. Try to play PA (or any kind) in a lobby game. Try to get a crit. Then count if the next attack is also a crit. If this article is wrong, the chance of next attack be a crit should be 15%, because each attack is independent event. If this article is right, you should see the chance of crit is lower than 15% (4%, if I'm not mistaken). The experiment has been repeated many times in history of dota (both 1 and 2), you can look for them.

    Dieser Kommentar wurde geändert
    John Silver (Donor)

      короче, как в покере, если кто-то понимает о чем я)

      t0kimonsta-

        @Mandalord
        Then I suppose this article is correct in explaining how the Dota 2 mechanic works, however I disagree with the way the 'pseudo' RNGs are coded in Dota 2 using counters. If the purpose is to 'manage luck' as you said, then the best way would be to decrease the variance of the Binomial random numbers simulated from a real RNG. In that way, we can stop this weird trick of predicting when you have the highest chance of crit-ing on PA, and make the critical events independent. There are methods to do decrease variances on random numbers using Antithetic variates, which in my opinion would be a more clever way of coding it.

        bones

          @Cunnilinguistitian I'm pretty sure he is just saying that in dota 2 the term "pseudo random" doesn't mean what it means in mathematics.

          Fake Xer0™ 様

            @Cunnilinguistician Ok. We all got your experiences with those random or pseudo-random, how educated and great you are. The point is, we gamers don't care all about those bullshit. We don't care true or pseudo. We only care if the facts can be used to improve our game. I think this article just served its purpose. Just stop making yourself a joke. We all had fun though.

            Abstract

              This has been in the game forever and I've never seen a case where someone used it as an advantage. Would be interesting to see if it could actually be done. As far as I can tell there are almost 0 situations where there is both usable time and usable health to sit in the jungle cancelling attacks on creeps to setup an early game gank, as I'm sure this strategy would be near pointless when late game items have been achieved. It's possible though.

              Dieser Kommentar wurde geändert
              Fixie

                @Cunnilinguistician
                The article doesn't refer to the generation of pseudorandom numbers in general. In dota 2 the word pseudorandom means that the actual chance to bash/crit whatever is slightly lower than the listed chance and increases with each attack that doesn't crit/bash. This is a system that makes no crit/crit streaks less likely while the average chance to bash/crit is close to the listed one.

                That certain random abilities work this way has been confirmed in patch notes when the system was introduced.

                Your comment is simply the result of misunderstanding the arguably confusing dota 2 nomenclature.

                Legendary NexO

                  Currently the Enchanted Mango only has a 31.43% win rate. Together, we as a community, have the ability to fix this problem and end this horrific tragedy. Please help. It is our duty as gamers to give this item the respect it is due . KAPPA

                  Keep Awake

                    @cunnilinguistician I don't know what you guys are talking about exactly, but I'm not sure how much you know about this game: there are two types of random distributions used in it. I have no idea how they work but i do know one is called true random distribution and the other pseudo random distribution.

                    La Dwin

                      Best article you guys at dotabuff posted. Learned a lot

                      Feel The Pasta

                        Changed top image 3 times XD

                        kurwa bober

                          hui zalupa chlen jopa ssanina pizda vlagalishe moshonka elda

                          Aerndalf der Ährenlose

                            It can be used very easily since most chances are fairly low and you basicaly have a really high amount of tries to stack them up available... the best example would be PA since you can stack with your autoattacks and proc with the dagger... with every item now too..

                            TicoKz

                              Шикарная статья, пойду юзать

                              fireattack

                                I didn't get the first graph. If it's true random, your probability of proc should be a constant 15%...

                                Schadenfreude

                                  PHANTOM LANCER´s juxtapose is pure random or is pseudo-random??
                                  i really want to know, i love this hero and it could help me

                                  Lord_DVD

                                    I tried this with PA and it works like a charm....

                                    Thank you. Wonder Cow

                                    Taco Cat

                                      Good Tip

                                      bogey654

                                        Luck should be reomoved from the game. 'Oh no, I lost 50 MMR because in 2 games I got chain bashed.' This shit skewes matchmaking.

                                        It won't be removed though.