Backgammon NJ: Our detailed responses to false statements posted on the App Store
Here are our responses to some of the false claims made by users in reviews for Backgammon NJ on the App Store. Since developers are not given a way to respond to reviews on the App Store, we're responding to them here.
You can also read about how to prove that Backgammon NJ doesn't cheat here.
| Posted by: |
(Various) |
| False Statement: |
(Note: Comments about reviewers who claim to be advanced players.) |
| Our Response: |
Some people who claim to be "advanced" or "ranked" players post false statements on the App Store. You'll find that these people have the following in common:
- They have not identified themselves by name. It's easy to post false reviews anonymously with no accountability. If they were truly ranked at world-class levels and could hold their own against the Expert level of Backgammon NJ, then they would be known in the top backgammon community. They wouldn't mind identifying themselves by name and posting to BGonline.org in order to prove their false accusations among their peers. If you'd like to read reviews for Backgammon NJ by actual world-class players who have identified themselves by name, click here.
- They have no verifiable ranking and no verifiable FIBS rating. Even if they are ranked in their local backgammon club or have beaten everyone they know in their personal circle of friends, that doesn't mean that they play at world-class levels and can hold their own against a top backgammon program. There are very few people in the world who play at such high levels.
- They don't know how to use other top bg programs to analyze matches or perform rollouts. If they knew how to fully use other top backgammon programs such as GNU Backgammon or Snowie, then they would export matches from Backgammon NJ and import them into other top bg programs. Doing this, they could analyze the luck factor for Backgammon NJ, see that it's honest, and also perform analyses to rate the skill level of Backgammon NJ.
- They don't recognize the strength of neural network bg bots. The top players in the world know the strength of neural network bg bots and use them to practice and improve their game. Only those without experience would claim that computer programs can't play better than humans.
|
|
| Posted by: |
(Various) |
| False Statement: |
(Note: Comments about doubling.) |
| Our Response: |
People who claim that the AI gets "luckier" after they accept a double are showing that they don't understand doubling strategy.
On Expert level, the AI will only double when it has roughly a 65-75% chance to win and redouble when it has roughly a 70-80% chance to win. The actual algorithm for how it determines when to double is much more complex. The exact detailed mathematical explanation can be found here:
http://www.bkgm.com/articles/Janowski/cubeformulae.pdf
A more general explanation of doubling strategy can be found here:
http://www.xfriis.dk/maxfriis/bg/double.html
Until you have a better understanding of when to accept or decline a double, we suggest that you follow the advice given by the Cube Tutor in the game.
|
|
| Posted by: |
(Various) |
| False Statement: |
(Note: Comments about statistics.) |
| Our Response: |
People who think that a 1, 2, 5%, or some other small percentage difference in any statistic between them and the computer is out of the ordinary have a lot to learn about statistics. If you look at a small sample size (for example, 200 rolls), stats can vary widely, possibly 50% or more from expected values. The larger the sample size, the more closely the statistics will approach expected values, but they'll never match exactly. For a closer approximation, you would likely need to look at a sample size on the order of 100,000 rolls or more. Go to wikipedia and read about sample size and statistics.
If you're always looking for things to blame your losses on, then your gameplay won't improve. But if you do what the experts do, change how you approach the game, and work to improve your moves by lowering your Error Rate and increasing your Checker Play Skill Percentage instead of focusing on win/loss, you'll find that over time, your game will improve, and as a result, you'll win more.
|
|
| Posted by: |
(Various) |
| False Statement: |
(Note: Comments about rolling doubles.) |
| Our Response: |
People who think that the game rolls too many doubles don't realize that doubles occur much more often than they think. The chance of rolling a double is 1 in every 6 rolls or 16.7% of the time. That means that if you play a game with approximately 50 rolls total for both sides, then about 9 of those rolls are expected to be doubles.
However, a single game of 50 rolls is a very small sample size. And when you look at a very small sample size, stats can vary widely from expected values. Thus, in any single game, it's very likely that you'll see more or less than the expected number of doubles. You'll need to look at a large sample size (on the order of 100,000 rolls or more over the course of many games) to get closer approximations to expected values. The larger the sample size, the more closely the stats will approach expected values, but they'll never match exactly. Go to wikipedia and read about sample size and statistics.
Backgammon NJ tracks the number of doubles rolled by both players under Menu: Statistics: Dice. There, you can quickly see your current sample size, the percentage of doubles rolled, and how close it is to the expected value (16.7%). As your sample size increases, your percentage will more closely approximate the expected value.
Backgammon NJ also saves match files for your last 10 matches under Advanced Options: Match Files. The match files show every single move and roll for both players for every game in the match. You can send the files to yourself and view them in a text editor such as Notepad or MS Word.
|
|
| Posted by: |
(Various) |
| False Statement: |
(Note: About manual dice.) |
| Our Response: |
Some people falsely state that the game doesn't include manual dice, where you roll real dice and enter the rolls into the game. This isn't true. Manual dice are included in every version of Backgammon NJ. To access this feature, scroll down in the Settings window and turn Manual Dice on. |
|
| Posted by: |
(Various) |
| False Statement: |
(Note: Comments about particular reviewers who post false statements on the Apple US App Store.) |
| Our Response: |
SwordShadow - This person has updated his review 40+ times. To read about him in detail, see below.
Girshon Rutstien/Music Wiz - This person updates his review constantly, probably 30 to 40 times or more, but we've long since lost count. Everything he complains about is countered by the dice stats in the app, yet he refuses to use them. He doesn't have the capacity to comprehend the proof we've given that shows that the computer doesn't look at upcoming rolls. He claims to play other top desktop bg programs, yet he has no understanding of how to use them. Just because one bg program plays a particular move differently than another doesn't mean that one or the other cheats. If they played exactly the same, then one would be a port of the other. We challenge this person: Post your last ten Backgammon NJ match files for all to see. Let the real experts analyze your matches, run them through GnuBG to determine luck factor, and truly see how well you play. We'll be waiting with bated breath for your post...
Sigea - This person has updated his review 13+ times. In one of his initial reviews, he claimed that the fact that the app didn't roll doubles during the opening roll was proof that it was cheating. When we explained that smart programs discarded doubles during the opening roll so as not to make the user waste time re-rolling, his statement saying that was proof of cheating conveniently vanished, without any admission of his own error.
Drez12345 - This person claims that he's not able to look at his dice statistics because he hasn't rolled 1500 rolls. He must have been in too big of a rush to write his false review because he failed to notice that the dice statistics are displayed after the sample size message is dismissed.
Erlock, Persue, Trojer, D_dresner, Styer5, Ciapp, Dj Blue 42, Flyboytim, Jkiy2356, Laguna7, Stone Kisses, YuraFrayman - You'll always find false reviews from these people near the top of the list.
|
|
| Posted by: |
SwordShadow |
| False Statement: |
(Note: This person updates his review multiple times per week and sometimes even multiple times per day. He has never reviewed any other app on the App Store except for ours, and he has been updating his false review constantly for a period of over two years. Look for his false review near the top on the Apple US App Store.) |
| Our Response: |
Update: We were finally able to get this person to send us dice statistics using a special method in the game. The dice stats he sent us included detailed information which allowed us to track exactly how he was manipulating the dice to skew the stats. In short, he was looking at upcoming dice rolls, and whenever a sequence of repeated doubles was about to come up, such as 3 doubles in a row, he would restart the game on an opening roll which would result in the computer getting the 3 doubles in a row. He didn't do this on every sequence, but he did it enough to skew the dice stats to look like the computer was getting more doubles and doubles in a row.
This just goes to show that certain people are intentionally lying about our game in reviews on the app store. We have no idea why someone would do this other than if they were a competitor, troll, just out of general maliciousness, or some other incomprehensible reason.
The best thing you, as a potential buyer, can do is to read our detailed explanation (http://www.njsoftware.com/note.html) about how you can use the technical features included in the app to prove to yourself that the game is honest and then think long and hard about whether you understand and agree with the proofs before you buy the app. The proofs are the only way that you'll be able to convince yourself without a doubt that the game is honest. If you don't understand and agree with the proofs, then don't buy the app. You'll be much happier if you choose a different game with an easier AI.
This person has changed his story so many times, it's comical. He's updated his review 40+ times. At first, he claimed that the computer was getting 3.6 times as many doubles as he was. He emailed us and insisted that he kept track of the count over 100 games. When we asked him to send us the seed and roll numbers of some games he was playing so that we could double check his counts, he never did. In fact, there was never any proof of his claims, only his continually updated review on the app store that stated it as "fact."
After we put in dice stats, he changed his story to "the computer rolls 17.5% doubles to my 14.4% (484 to 398)". So where was this supposed 3.6x disparity? According to his earlier story, the computer should have had over 1000 doubles. For many months, he kept updating his review claiming the 3.6x doubles disparity. You have to wonder why he originally claimed the 3.6x doubles: did he miscount, or did he just lie?
This person claims that he is an expert with 37 years of backgammon experience. He sent us the following position as "proof" that the computer cheats, because it "had no reason to make the move it did". In this money game with no Jacoby, the computer (blue) played this 6-4 by taking a piece off and then hitting the piece on 2.
Upon analysis of this position, we determined that the computer's move was indeed the correct move and that the move this person claimed was better (taking two pieces off) was in fact a blunder. The computer's analysis was as follows:
6/2* 6/off, Win = 93.6% Gammon = 77.0% Backgammon = 1.3%, Equity = 1.66
6/off 4/off, Win = 92.3% Gammon = 66.5% Backgammon = 1.5%, Equity = 1.53 (-0.13)
We also performed a rollout in GNU Backgammon with this result:
6/2* 6/off, Win = 95.1% Gammon = 77.2% Backgammon = 0.9%, Equity = 1.72
6/off 4/off, Win = 91.6% Gammon = 67.9% Backgammon = 1.4%, Equity = 1.56 (-0.16)
With all this evidence, this person still claimed that no human professional would ever play the hitting move. So we posted this position to bgonline.org, where professional backgammon players go to discuss backgammon. So far, every expert on bgonline has agreed with the computer's play. One expert gave the following explanation:
- "I'd play 6/0 6/2*, for two reasons: 1) Wins, and 2) Gammons.
- If White rolls a 2, Blue would rather be hit and have to bring his checker around again (while hitting White off the 2pt repeatedly if he has to) than to have White make his 2pt. If White makes the 2pt, he can sit there and wait for Blue to leave a shot (If I recall, Blue does so about 70% of the time) while building his board for containment. It's much safer to just hit White and close him out and be done with it. I'd make this play at DMP. The gammons extra gammons are nice, big bonus."
Here is the link to the post:
http://www.bgonline.org/forums/webbbs_config.pl?noframes;read=63804
|
|
| Posted by: |
-- |
| False Statement: |
"The glowing reviews are paid plants." |
| Our Response: |
The good reviews are written by real users who we have no association with, some of whom are world-class players who have identified themselves by name in their review. If you really think that a world-class player who has competed at the World Backgammon Championship would jeopardize their reputation by identifying themselves by name and lying about a $5 backgammon app for money, then we have a bridge we'd like to sell you... |
|
| Posted by: |
-- |
| False Statement: |
"I don't care how many paid "giants" say otherwise." |
| Our Response: |
A Giant of Backgammon is one of the top 32 players in the world. If you can't believe a Giant, then who can you believe? And once again, if you think a Giant would lie about a backgammon app for money, then...see our response above. Better yet, go to the BGonline.org forums and ask them yourself. |
|
| Posted by: |
-- |
| False Statement: |
"...ai looks ahead at upcoming rolls and plays accordingly." |
| Our Response: |
To prove to yourself that the AI does not look at upcoming dice rolls, follow these steps. |
|
| Posted by: |
-- |
| False Statement: |
"I am considered expert level and have been playing backgammon for [insert suitably large number here] years and this app's "lucky rolls" make me want to throw this across the room." |
| Our Response: |
How long you've been playing backgammon has no bearing on how well you play. Some of the top backgammon players in the world right now are very young. If you're an expert player, then you won't mind proving the false statements you've made about our app with concrete evidence. Take the proof that we've provided (RNG Seed, printouts of dice rolls, detailed method that shows the AI isn't looking at upcoming rolls, etc.) and prove that it's not true. Post your evidence and your Backgammon NJ match files for all to see at the BGonline.org forums, frequented by many of the top players in the world, start a discussion there, and find out what the true experts think. We'll expect your post there any day now.
NOTE: We first issued this challenge on 2/11/10. To date, no one has posted to the forum. It's easy to falsely slander us with no accountability, but when asked to prove your claims in a forum of experts, you can't do it.
|
|
| Posted by: |
-- |
| False Statement: |
"Dice stats prove cheating... The random dice favor the computer..." |
| Our Response: |
First, dice statistics can be modified by using Manual Rolls and by changing the Next Roll Number and RNG Seed. Dice statistics are pretty charts, but they don't prove anything. The only way to prove your false claims is to take the detailed proof that we've provided that shows the game isn't cheating and prove that it's not true. See our previous response above for details on what you need to do.
Second, if you think that random means you're going to roll exactly the same number of doubles or other rolls as your opponent, then you have a lot to learn about statistics. The larger the sample size, the more closely the statistics will approach expected values, but they'll never match exactly. For a closer approximation, you would likely need to look at a sample size on the order of 100,000 rolls or more. Go to wikipedia and read about sample size and statistics.
Third, if you think you're right, then surely you're not afraid to post your match files. Post your last 10 Backgammon NJ match files for all to see at the BGonline.org forums, frequented by many of the top players in the world. The match files document every single move and roll made by both you and the computer in each match. Let the real experts analyze your matches, run them through GnuBG to determine luck factor, and truly see how well you play. We're waiting for your post...
|
|
| Posted by: |
-- |
| False Statement: |
"I've also noticed that if you have the "Auto Bear Off" on that sometimes it will move a man instead of taking one off." |
| Our Response: |
In many situations, it's correct to move a checker instead of taking one off. Backgammon NJ uses a pre-computed bearoff database, so it plays nearly perfectly in bearoff situations. You should note that there are 2 settings for Auto Bearoff: Greedy and Best Move. Greedy will always attempt to bearoff as many checkers as possible, even when doing so isn't the best move. However, since you have it set on Best Move, it will always make the best move, which sometimes means not bearing off a checker.
If you think that the computer made the wrong move during bearoff, then take a screenshot of the board and send it to us. We'll analyze it and tell you why the move it made was the best move. |
NJ Apps
|