Community RPG Codex 2014 Role-Playing Game of the Year Awards

[toro]

ofcurse, the tags - the last resource of any imbecile who looses an argument based on logic and common sense.

 

[toro]

Funny, look who's talking.

Point taken.

But your posts were even more filled with shit. I guess you have nothing to say about that?

 

[Vikter]

Lords of the Fallen in any top 10 list just makes me confused. If you're going to list something like that in 10th place, just make a top 9 instead.

 

[TwinkieGorilla]

hiver has appeared like the lone rider dumbfuck of the apocalypse to signal the end of this thread's relevance.

 

[Rake]

hiver has appeared like the lone rider dumbfuck of the apocalypse to signal the end of this thread's relevance.

This tread's relevance ended around page 5

 

[felipepepe]

However this ranking is done under the assumption that all votes are positive.

In reality one voter which ranks D:OS with 5 and W2 with 3 means that he actually voted negatively for W2 with -2 points.

From the data set it is possible to identify how many votes and points were negative.

Then the average score formula for each game can be written like this (positive points + negative points)/(positive votes + negative votes).

Based on this formula, the top would look like this:
1. Divinity: Original Sin 2,714630225
2. Shadowrun: Dragonfall – Director's Cut 2,397306397
3. NEO Scavenger 2,246621622
4. Wasteland 2 2,102318145
5. Legend of Grimrock II 2,059590317
6. Valkyria Chronicles 2,028436019
7. Legend of Heroes: Trails in the Sky 2,01369863
8. Heroine's Quest: The Herald of Ragnarok 1,996254682
9. Tales of Maj'Eyal : Ashes of Uhr'Rohk 1,96097561

And you know what: This actually looks good. No fucking m values and shit.

It took me a while to reply because I was thinking about it, and honestly I still haven't understood what you are saying... 1 & 2 are negative votes, 3 is neutral and 4 & 5 are positives. You treat only 5 as positive?

 

[Stelcio]

Point taken.

But your posts were even more filled with shit. I guess you have nothing to say about that?

Hard to make a sensible counterargument to rants and insults. You were the one who moved the discussion to this shit level. And once you started to make any sense, responses got better as well. Who would have thought, right?

 

[toro]

It took me a while to reply because I was thinking about it, and honestly I still haven't understood what you are saying... 1 & 2 are negative votes, 3 is neutral and 4 & 5 are positives. You treat only 5 as positive?

I did a mistake in the previous post therefore please ignore it. Also my approach has nothing to do with "1 & 2 are negative votes, 3 is neutral and 4 & 5 are positives".

We can all agree that the average score for a game is not the best way to sort the games: it basically doesn't say anything about how the game relate to the rest of the games.

Therefore in order to sort the selected games we need to compute how each game relates to the rest of the games from that specific set. There isn't only one way to do this.

Bayesian average is equivalent to taking the average score of a game and perform a weighted arithmetic mean with the average score of the entire set. This method is based on an assumption and all the average scores are normalized towards the set average.

My approach is equivalent to taking the average score of a game and perform a weighted arithmetic mean with the negative score of the game. This method is based on the data available and there is no normalization.

The negative score of a game is defined as the sum of all the negative votes for that game where the negative vote is relative to the max vote.

Example 1: If the voter has voted 4 for D:OS, 2 for W2 and 3 for Sdwr then his vote can be divided like this:
- positive vote 4 for D:OS, positive vote 2 for W2, positive vote 3 for Sdwr.
- negative vote 2-4 (-2) for W2, negative vote 3-4 (-1) for Sdwr. D:OS has no negative vote here.

Example 2: If the voter has voted 5 for W2, 1 for D:OS and 2 for Sdwr then his vote can be divided like this:
- positive vote 5 for W2, positive vote 1 for D:OS, positive vote 2 for Sdwr.
- negative vote 1-5 (-4) for D:OS, negative vote 2-5 (-3) for Sdwr. W2 has no negative vote here.
---------------------------------------------------------------------------------------------------------------------------
D:OS - 2 positive votes: 4, 1 and 1 negative vote of -4
W2 - 2 positive votes: 2, 5 and 1 negative vote of -2
Sdwr - 2 positive votes: 3, 2 and 2 negative votes: -1, -3

On IMDB one voter can vote for only one movie at a time while a vote on your survey is equivalent to multiple votes. Because of this, your data set has additional information.

Bayesian approach considers all votes as positive (how much the game is liked) while my approach also considers the negative votes (how much a game is hated or is not good as).

Bayesian average formula is: WR = (set average score * min votes + positive average score * positive votes) / positive votes + min votes

My approach formula is: WR = (negative average score * negative votes + positive average score * positive votes) / positive votes + negative votes

For something half-arsed, the results are quite good:

1. Divinity: Original Sin 2,714630225
2. Shadowrun: Dragonfall – Director's Cut 2,397306397
3. NEO Scavenger 2,246621622
4. Wasteland 2 2,102318145
5. Legend of Grimrock II 2,059590317
6. Valkyria Chronicles 2,028436019
7. Legend of Heroes: Trails in the Sky 2,01369863
8. Heroine's Quest: The Herald of Ragnarok 1,996254682
9. Tales of Maj'Eyal : Ashes of Uhr'Rohk 1,96097561
10. Lords of Xulima 1,691304348

These results are equivalent to the Bayesian average when m is between 15 voters and 50 voters. Which is much better than going with the min 7 voters.

I might be retarded but indulge me. Take a look and tell me why this is incorrect.

 

[Viata]

Nah, I still like Heroine's Quest being above Wasteland 2.

 

[Stelcio]

I actually see it making sense, toro. If I interpret the formula correctly it basically favors games that were rated highest amongst one voter's votes by pulling down the ratings of games that may have been rated moderately high, but still lower than others. That's a nice idea indeed.

 

[toro]

I actually see it making sense, toro. If I interpret the formula correctly it basically favors games that were rated highest amongst one voter's votes by pulling down the ratings of games that may have been rated moderately high, but still lower than others. That's a nice idea indeed.

You are correct on this regard.

Edit: But I would not call this method as favoring any set of games: the same computation is applied to all of them.

 

[felipepepe]

I still need to reply to your new methodology, but I'm out of town in a PC without excel, and I HATE using google docs.

 

[tuluse]

On IMDB one voter can vote for only one movie at a time while a vote on your survey is equivalent to multiple votes. Because of this, your data set has additional information.

I don't see how this is true. What's the difference between voting for 50 movies on imdb in a row and having them all on one form?

Edit: Also, I was playing around with the m value in bayesian, and my initial hunch that 50 was a good number turned out to be false I think. With an m that high, no game was rated under a 2. I then lowered it to 20 and we had two games under 2. So I think 20 gives better results. I do still think 7 is too low.

 

[toro]

I still need to reply to your new methodology, but I'm out of town in a PC without excel, and I HATE using google docs.

No problem. When you have time.

I don't see how this is true. What's the difference between voting for 50 movies on imdb in a row and having them all on one form?

You are right, the same information can be extracted after the voting is over therefore there is no fundamental difference.

And yes, m seems to be between 15 and 45 which would put Wasteland 2 on position 4.

 

[coffeetable]

Because is equivalent to state that the opinion of 7 voters for Legends of Persia has the same weight as the opinion of 906 voters for Divinity:Original Sin.

what

okay, look. here's the likelihood & posterior for LoP and DOS when a uniform prior with pseudocount m = 7 is used

note how LoP is dragged towards a flat distribution, while DOS is hardly affected

 

[agris]

felipepepe in regards to analyzing the relationships between votes, check out principal component analysis. It's one of the most rigorous analytical methods to examine correlation between variables that I know of.

 

[felipepepe]

Ugh, I'm really not in the mood for math atm... but the excel sheet is on the first page, so feel free to give it a try.

 

[Bubbles]

Neo Scavenger and RPG Codex Goy 2014 (Rant)

Only now in retrospect to I realize what an incredible year 2014 was for pc rpgs. Divinity:Original Sin, Wasteland 2, Might and Magic X, Lords of Xulima, Valkyria Chronicles, Legends of Heroes: Trails of the Sky part 1, Shadow Run: Dragonfall, Legend of Grimrock 2, and Dragon Age:Inquisition (which is the only one I still haven't played because it runs too slowly on my computer). Also Neoscavenger, which I've finally got to playing after I got it on the GoG summer sale. Neoscavenger won 3rd place on the RPG codex GoY award. It just boggles my mind. Before LoX, MMX, VC, LOH, LoG, or DA:I. How can that be? Don't get me wrong its a fun little game with pretty intricate mechanics and unlike most roguelikes, surviving disease, hunger, and freezing is more important than combat, but in terms of combat, story, scope, and content its way behind every other game on that list. So how could it reach 3rd place? This is the problem when like 8 people are super fanatic about a game, and their votes outweigh when 400 people choose something as GOY and 100 people don't like it. I mean its nice in that it might point people to game they might not otherwise play, but it falsely puts games which really appeal to only a very small group over something that most voters like. I just have big problems with this kind of voting system. (rant out).

http://www.rpgwatch.com/forums/showthread.php?p=1061410496

 

[vonAchdorf]

Neo Scavenger and RPG Codex Goy 2014 (Rant)

:D

Before LoX, MMX, VC, LOH, LoG, or DA:I. How can that be?

I'm glad that the 2014 voting / weighting system hasn't been discussed to death already.

 

[Infinitron]

Goy

>_>

 

[Lhynn]

Neo Scavenger is better than all those shit games. xulima was shit and boring, MMX was mediocre, Valkyria Chronicles had a female protagonist, LoH is as generic jrpg as they come, DA:I was complete wank.

 

[Zorba the Hutt]

I like the way he admits he hasn't played DA:I on his toaster yet, but is still butthurt that another forum voted for Neo Scavenger ahead of it.

 

[Perkel]

Neo Scavenger won 3rd place because it had pretty much everything better than those games mentioned. Innovative take on turn combat, great lore and world design, ton of interesting systems to play with like tracking that actually works, hiding your smell from rapewolves who follow you, having diahrea from eating shit food, ton of small situatutions that made you think constantly if just another scavenge is worth risking your life. Ton of intersting skills which actually gave you something that had worth in that world. Interesting NPCs and locations.

What is best replayability. You could start new build focused in different way and you would have different playing game.
It was game which was substance over style.

 

[Viata]

Well, two years later and this awards still causes some butthurt.

 

[CryptRat]

Valkyria Chronicles had a female protagonist