Why The Stats Are Bs....

[Bezo 0]

I apologize if something to this effect has already been posted before. I was just wondering if anyone else thought the stats kept on the Lotus site are BS. As you can all see, my stats are pretty shitty. My K/D ratio is less than one and my kill frequency is not very high. However, I frequently lead in points on the 32-man dustbowl and goldrush servers, and I am almost always one of the 3 MVPs if my team wins. Why is this? I play mainly medic, engineer, and spy. These classes all get the screwjob on the stats server since some of their main duties are not taken into account whatsoever.

I do think that the stats are a general indicator of skill, and no one will deny that top players on the stats server like killerkeg and honky doomsday are the best players...but for a lot of people, the stats do not accurately reflect skill level. Since I am one of those, the stats are BS.

[Cheese-kun 0]

True stats are a semi-good indicator, but don't let it actually get to you. I mean all I think it does is inflate egos really. Like for example, that american ninja chuck norris guy tried to voteban me from lotus because I sniped him twice, and according to him since he's so good, a guy named cheese has to be shit compared to him and therefor I must be hacking >_>. Anyways don't let stats get to you, all that should really matter is having fun not some numbers

[You suck and have no life 203]

Stats are very good indicator of skill. Everyone on top 100 list, 90% of the people are very good players and thats good enough for me.

Just like the SATs or any standardized measurement tests, they arent a 'perfect' measure everything, but if someone got a 1600, id put my dollar on him that hes smarter than someone with a 800 score, but hey, I could be wrong, maybe 0.001% chance.

You have to look at the big picture here. There are 100,000+ people who have played on the server, the stats may not be accurate to +/- 1 rank, but even at +/- 1000 rank, its still accurate to 1% of 100,000.

The stats is simply a set of equations with coefficients and multipliers. You do a specific task, you gain a certain amount of points, you die, you lose a certain amount of points. People who are good tend to do more of the former than the latter, hence more points overall. Doing certain things gives your more points, and those multipliers are man made. They arent assigned by god. So if you think the rank system is BS, ignore it. On the flip side if you want simply a high rank, look at the formula for how the points are added, and simply do what gives you the most points. I will tell you one thing, k/dr is a big factor on how its calculated. A k/dr of 1 will put you pretty low. a k/dr of 2 will get you about a rank of 100. A k/dr of 3 will get you approximately 30+ ish (all appoximations).

[SoWz 0]

I think the stats were originally created for games like counter strike, where there are no different classes.

The skill level only screws over medics, because they usually have no kills and tons of assists (but your skill still goes up).

If you get a lot of points as a spy or an engi, then you should have a lot of kills, which will be reflected in the skill level. The medic is probably messing up your K/D ratio.

[DarkClaw 0]

  Stats are very good indicator of skill. Everyone on top 100 list, 90% of the people are very good players and thats good enough for me.

Just like the SATs, it doesnt measure everything, but if someone got a 1600, id put my dollar on him that hes smarter than someone with a 800 score, but hey, I could be wrong, maybe 0.001% chance.

You have to look at the big picture here. There are 100,000+ people who have played on the server, the stats may not be accurate to +/- 1 rank, but even at +/- 1000 accuracy, its still accurate to 1%.

Ok, what? I didn't understand that... in english next time plz

And no... not just medic. It's one of the spy's job to bring the intel out into the open if it's too heavily buried under the team's defence, or to start capping an unguarded point (like the before last point in dustbowl... I usually have to go spy to cap that one). Usually though, you get killed before you get to your intel room/finish capping, and someone else gets the point instead and get no points whatsoever... so in the end, your only worth what you're worth... if you find your doing a good job and helping the team, good for you, don't mind the stats, they won't record things like "Sacrificing one's self to help team"

[You suck and have no life 203]

Ok, what? I didn't understand that... in english next time plz

7906[/snapback]

[Alucard winrar 0]

I find stats to be a good indicator of who is a good fighter or killer but it doesnt say much about support classes mainly the medic. I just dont worry about it, stats are just numbers on a website and while its nice if your in the top 100 it doesnt do you any good or give you a prize. Honestly some ppl worry about stats to much and it interfers with the point of the game TEAM fortress 2(so do achievements) where its nice if you have a good k/dr but sometimes you have to stick your neck out to push your team onward.

[You suck and have no life 203]

There IS correlation between skill and rank. If it didnt, then a good player is just as likely to be last ranked as he is first ranked. Then it would be graph 1, which is not the case.

A perfect correlation, would be a linear line with an R2= 1, graph 3, which is also not the case. It is impossible for any rank system to be this good at assessing peoples true skill level, since the rank is simply determined by points, which is calculated by simple equations.

Graph 2 represents a real life, typical ranking system. The numbers are made up but its to make a point. The correlation is not super tight, and a good player may be ranked slightly lower or higher depending on situation, what class he plays, etc etc. Take me for example. I fluctuate between rank of 33 to a rank of 300.

[coda 1]

the way skill works is it looks at the skill of the person you kill (or kills you) and generates a probability. for example, if you have a skill of 1600 and you kill someone with a skill of 300, psychostats will think "easy kill, i'm not moving up your stats". on the other hand if you have a skill of 300 and you kill someone with a skill of 1600, psychostats will think "omfgbbq, +100"

hopefully that'll help

[friedrice 8]

I play mainly medic, engineer, and spy.  These classes all get the screwjob on the stats server since some of their main duties are not taken into account whatsoever.

7900[/snapback]

I don't see how you can have such a bad k:d ratio if you play a lot of engineer and medic. The simple fact is that if you die a lot as either two, your doing something wrong. As a medic, it is better 90% of the time to go get health rather than let yourself die. Especially on maps with a longer respawn time.

As an engineer, you are almost never a target, your sentry is. If you see a bunch of stickies landing near your sentry, RUN! You will rebuild much faster if you don't have to wait to respawn.

I do see your point with the spy, especially since the stats don't keep track of destruction. But in any case, if you die more than you kill, your not doing a very good job. I think Spy is one of the harder classes because so many people try to backstab the first person they say, or run to the closest sentry.

To agree with most of the people who have posted, the stats are not perfect, but more of just an indicator, as well as for fun. Besides, you should learn from what the stats tell you and figure out what your best at and what kills you the most.

[Bezo 0]

I think that my main issue is with destruction. For example, I mostly played spy during my last dustbowl round. On the overall points list, I was 40 points ahead of the closest person on either team with about 150 points. I don't remember the exact stats, but it was something to the effect of:

44 kills

50 deaths

48 assists

52 destruction

5 CP

On almost every round, I cleared many SGs/dispensers and most of the teleports. IMO clearing 1 teleport exit is = to 1 kill or more...especially if the action is far from the opposing team's spawn point. At the very least, destructions should be given half the weight of a kill on average. After all, killing someone (especially if you are on the defending team) is no great accomplishment if they can respawn and teleport back in 5 seconds.

[coda 1]

I think that my main issue is with destruction. For example, I mostly played spy during my last dustbowl round. On the overall points list, I was 40 points ahead of the closest person on either team with about 150 points. I don't remember the exact stats, but it was something to the effect of:

44 kills

50 deaths

48 assists

52 destruction

5 CP

On almost every round, I cleared many SGs/dispensers and most of the teleports. IMO clearing 1 teleport exit is = to 1 kill or more...especially if the action is far from the opposing team's spawn point. At the very least, destructions should be given half the weight of a kill on average. After all, killing someone (especially if you are on the defending team) is no great accomplishment if they can respawn and teleport back in 5 seconds.

7912[/snapback]

So you're best at killing inanimate objects?

btw, guys read my above post, it's not just about kd

[Bezo 0]

Right...it's much easier to kill people with spam crit rockets and demo bombs than to run invis through a team of trigger-happy pyros to get behind sgs...then take out the sg, dispenser, and engi all at once.

In any case..I think you're missing the point. The point is that destructions (not just from spies) and healing play a large role in all maps, and to leave out those stats completely does not give a full picture of skill. That's all.

[DarkClaw 0]

There IS correlation between skill and rank.  If it didnt, then a good player is just as likely to be last ranked as he is first ranked.  Then it would be graph 1, which is not the case.

A perfect correlation, would be a linear line with an R2= 1, graph 3, which is also not the case.  It is impossible for any rank system to be this good at assessing peoples true skill level, since the rank is simply determined by points, which is calculated by simple equations.

Graph 2 represents a real life, typical ranking system.  The numbers are made up but its to make a point.  The correlation is not super tight, and a good player may be ranked slightly lower or higher depending on situation, what class he plays, etc etc.  Take me for example.  I fluctuate between rank of 33 to a rank of 300.

7909[/snapback]

Man, I'm not in university yet (although studying engineering was something I'm thinking of)... Although I do get most of what your saying now... but anyways, what I'm arguing is that stats aren't good at assessing a person's REAL skill. I play spy, yet I focus more on destructions and caps. Do I get any good points with the stats system? No. But I know I'm doing my job to help my team, and I get in the top 3 players a lot of times. My rank is shit, I don't care. That's all that really matters, but if your one to strive for perfection and such, go ahead and try to get to the top rank. I'm just saying... any person who does their job but doesn't get too many kills is just as good as someone who gets a lot of kills... maybe even better if that person doesn't even do their job!

[Mana 3,867]

Ubercharges, destructions, captures, assists, etc all give points. Some of these give as many points as 3-5 kills do...

[DarkClaw 0]

Yes, but getting a kill is much easier to get than a cap/uber/destructions (well, when it's being guarded, which is basically most of the time), thus basically you still lose in a way... And many people forget to pay attention to their medics, thus letting them die easy, so maybe not even getting 1:1 Assist/Death ratio, since assisting also relies on the person you're helping. It may seem like it evens out because of the point increase, but in most cases I've seen, it doesn't. Everything other than kills are actually more difficult to get.

[Novas 0]

Stats are meaningless tba

[Honky Doomsday 0]

Its really all pretty pointless, as it was said, these stats were created for games like CS and DoD where your K:D ratio is plastered on the scoreboard for all to see and where your points really matter. If you care about stats you should just play demo or soldier and get a boyfriend to follow you around as a medic with kritzkrieg everywhere, but don't forget to run back to your spawn to charge up again making sure you don't die! Seriously though, stats in TF2 are pretty useless, its all about winning the round and playing like a team. Points do nothing in TF2, unless they are Control Points, in that case fucking STAND ON IT!

[Doxt 1]

stats in TF2 are pretty useless, its all about winning the round and playing like a team.

7937[/snapback]

couldn't agree more

[Josh 18]

couldn't agree more

7939[/snapback]

also agree, plus when I use to play alot the stats were down and they never registered my scores, I have at least 150 more hours then it shows on lotus...

[Mana 3,867]

Rounds wins give points, round loses take away points.

It's much more than just Kill Death Ratio. PS for CSS uses something totally different than PS for TF2.

These stats are just as meaningless as your Score on the scoreboard. TF2 calculates your Kills, assists, captures, defenses, flag caps, etc etc. So does PS. If there is a function you see in game that gives scores you can best bet TF2 logs/records it and PS counts it.

[Josh 18]

In reality, your score means nothing, its how you work for your team, how you help win the match, some people camp in the back and pick off people; others go and do the dirty work by going behind enemy lines. ^^

The life of a spy ^^

[You suck and have no life 203]

I dont think stats are meaningless at all. You form a team with the lowest ranked 6 players, and ill form a team with the highest ranked 6 players (yes including solahma) and well see how many games in a row my team will win. 100-0? 1000-0 perhaps?

You are never going to have a perfect program that will rank everyone perfectly, its simply impossible. In the real world, you aren't going to find a lot of things accurate to 100%; that doesn't make it BS. Lets say for example, if you are in the stock market, and you are right 80% of the time, does being wrong 20% make you BS? I would hope not because if you were right 80% of the time, you would be far richer than Warren Buffet.

There are a lot of variables including random ones the program will never be able to calculate.

For example:

If you suck, and its because you are drunk, the program doesnt know any better! If you got your ass kicked by a pro level competitive player, but he is new to the server and he has a low rank, the program doesnt know any better! That player will drastically pull your score down, because the program thinks that player is newb, and since you are losing to a newb, you must be a newbier newb! You want the software to be an all knowing omniscient god? It aint gonna happen.

I am a man of statistics. If a rank system is 90% accurate, that's good enough. It someone is ranked 1000+ higher than another person, I would bet my apples hes probably better.

People always come up with anecdotal examples, or select few situations here and there (minority) to make an assumptions or contradict the majority. Example: "well so and so could do this and this or blah blah to inflate his score". Yes, that's true, but most people DONT, therefore the rank system is representing the majority. Its like saying "well my grandma has been smoking for 30 years and shes 90 and still alive, therefore smoking isnt bad for you at all".

Using anecdotal examples to make a generalization is very stupid. In statistics, this is making a claim with insufficient sample size. Its like taking a poll with 1 vote, and making a conclusion saying that 100% agree one way. If your sample size is too small, your conclusion is weak.

For example:

Spy Enemy Disguise vs Friendly Disguise. If friendly disguise gives you an overall success rate of 60% and the Friendly Disguise gives you a success rate of 40%? Which is better? Obviously the Enemy Disguise is in general better. You cant say the Friendly Disguise is better because it works in certain situations even if it "could" in fact be better in certain situations. You cant predict future situations and you dont know what situation you are going to be in, therefore, in OVERALL situations, its worse off. Maybe if you wall hacked it might work. But my point is, you cant use a few examples to justify the big picture. If you want to make a generalization, at least 50% (if not more) has to agree with you.

[Josh 18]

pshh enemy disguise all depends on the spy, spy is 50% luck and 50% skill

if you know how to play spy then there is a good chance they won't touch you but that other 50% of luck is where they do a random spy check or someone runs some random ass way and hits you when invis. thats when the skill comes back in and determines if you can run away to safety in time but at the same time that luck factor plays back in on, how many saw you, how much time do you have left of invis, ect

[You suck and have no life 203]

When you compare two things, you need to keep other variables constant. If you dont, you are adding noise and you dont know if your results are caused by the things you are comparing, or due to the other variables.

For example:

If you want to compare two cars, and their gas efficiency. Well obviously when you test both cars, they should have the same wind resistance, incline, temperature. If you test one car going up a 20% incline hill, vs another car going down a 20% hill. You are obviously screwing up the experiment.

Example 2.

You want to test whether whey protein or casein protein makes you stronger. You cant use all bodybuilders in one group, and a buncha scrawny tf2 nerds in another group because now you wont know if the strength is coming from the type of protein, or is it just coming from the person who happens to be naturally stronger.

This applies to video game dynamics. When you say which is better, you most hold ALL other variables CONSTANT. This of course is not 100% possible, things fluctuate, that why you play over an extended period of time and take an average.

When you are talking about friendly disguise vs enemy disguise, you have to do the same thing. The player's skill level must be constant, and the opponent skill level must be relatively constant. I can assure you, if you played over a month using only enemy disguise vs a whole month of friendly disguise, your rank, kdr, score, and everything will be higher when using enemy disguise.