Edit: And if you want to know a little more about some other stuff I did alongside this one, you can see the results here: https://www.steamgifts.com/discussion/EWTuj/so-about-this-whole-reading-descriptions-thing-another-experiment

Last year I did some experiment by posting 12 copies of the same game, going form level 0 to level 10 in the process, to see how entry numbers will fare.
This year, I did it again, but this time I spiced things up a bit by mixing in quite a few other games into those HOGs. These included games from OtakuMaker or directly from the DIG point store.

Some of these games were not on the bundle list when I posted the giveaways. Some were, but never went into any bundle, it was just the usual Russian prices pushed it over 95% case. All of them, including some of the HOGs, are borderline or flat-out shovelware at worst, cheapo indie games at best. And this is how they went:
(Sorry about the formatting, the site doesn't support table column width in markdown.)

Title Bundled Rating (*=cards) 0 1 2 3 4 5 6 7 8 9 10
Crystal Cosmos 3 100%* 2305 - 854 - 909 - 562 - 182 - 56
Green Ranch 4 87%* 1900 1082 678 741 730 572 447 295 134 101 -
Sixtieth Kilometer 7 78%* 1074 - 438 - 408 - 160 - 39 - 6
Enigmatis: The Ghosts of Maple Creek 6 84%* 1524 909 485 413 344 196 139 73 37 19 15
Enigmatis 2: The Mists of Ravenwood 10 97%* 1272 803 406 330 259 119 63 22 8 6 3
Enigmatis 3: The Shadow of Karkhala 2 93%* 1804 1428 926 865 790 609 463 316 178 111 56
Bold New World 0 23%* 1881 1546 814 785 751 573 432 320 190 120 64
Pester 15 60%* 1332 1093 304 368 348 241 166 108 69 41 16
[Dream Dealer △](Dream Dealer △) 0 85%* 1988 1769 932 900 840 664 507 357 206 147 73
Blaite 1 77%* 1447 1296 742 732 732 585 460 341 208 140 68
POLYWAR 0 79% 1468 1109 526 518 511 414 313 235 137 97 51
Joana's Life 1 35% 1580 1241 582 511 496 427 325 241 144 102 49
City of Chains 4 86%* 1178 1003 437 374 334 230 149 71 36 20 10
Grim Legends: The Forsaken Bride 11 95%* 1054 919 314 210 123 45 14 2 0 0 0
Grim Legends 2: Song of the Dark Swan 8 91%* 1288 1041 421 314 223 103 45 16 2 4 2
Grim Legends 3: The Dark City 2 92%* 1816 1575 837 739 642 490 373 261 148 - 45
Demon Hunter: Chronicles from Beyond 7 64%* 1066 680 397 322 283 187 109 63 23 17 6
Demon Hunter 2: New Chapter 2 67%* 1224 889 576 513 504 406 327 242 148 98 50
Control Craft 3 2 60%* 1300 745 465 426 424 358 299 226 142 101 51
Finnish Roller 3 75% 1347 830 486 449 449 402 345 269 171 118 -
Infinitum 2 25%* 1527 751 452 416 413 364 330 244 157 111 56
Trashville 0 88%* 931 757 429 451 456 342 292 214 129 82 40
The Secret Order 2: Masked Intent 5 79%* 860 814 421 378 332 217 172 127 73 50 18
The Secret Order 3: Ancient Times 3 90%* 910 887 484 490 412 295 247 189 114 75 36
The Secret Order 4: Beyond Time 1 82%* 1019 903 532 521 485 381 319 237 145 100 48
Oldschool tennis 1 50% 1080 805 397 396 388 311 270 182 121 89 41
The Lords of the Earth Flame 2 65%* 1025 827 415 432 421 343 275 189 108 72 32
9 Clues: The Secret of Serpent Creek 8 90%* 744 710 259 246 217 130 81 40 19 11 5
9 Clues 2: The Ward 5 93%* 1149 892 423 377 301 206 149 102 67 41 16
Purple Hills 1 40% 2067 1346 594 508 481 367 287 191 111 80 40
Agent Walker: Secret Journey 1 69%* 2145 1611 880 799 758 597 461 335 206 148 69
Naval Warfare 12 53% 1226 843 369 350 342 222 174 117 66 37 15
Endless Fables: The Minotaur's Curse 1 96%* 1911 1328 834 851 805 610 497 342 205 141 68
Crime Secrets: Crimson Lily 3 67% 1883 1325 573 417 391 333 267 194 113 72 31
Turn Around 1 45%* 1443 1070 531 540 526 436 384 286 173 119
Sarab: The Dark Tower 1 100% 1803 1268 490 463 421 315 277 198 129 87 42

Note: The table is not complete, I still have games left to add more data.

There are a few things to point out:

  • Low levels are completely chaotic. Sometimes there is a giant gap between level 1 and level 2 (look at Pester), sometimes the curve is almost ideal, sometimes there are more entries on level 3–4 than on level 1–2.
  • Game rating is almost irrelevant on public giveaways. Be it great or a pile of shit, there is not enough difference to draw a direct cause and effect conclusion between game rating and entry popularity.
  • Bundled status obviously has an effect, but only a small one on level 0 and level 1. This mostly indicates that those levels have most of the users who enter anything for a win or who use some automated methods to constantly spend their points.
  • On the other hand, considering the not-so-great quality of many of the games, we can safely say that people who don't really care about anything but a +1 in the library are common enough on all levels. Almost one-third of the level 10 users entered for Bold New World, one of the universally lowest-rated games on the list.
  • Additionally, bundled status won't automatically mean or predict anything. A similar amount (or, on some levels, more) people entered for Pester, a game bundled fifteen times, as for City of Chains, which was bundled about a quarter as often.
  • Level 6 and on seems to be the point where users start to actually consider what to enter for or have enough games in their library to not see many giveaways. Still, the allure of a +1 is obviously strong.
  • Level 8–10 seems to exist in its own world. Sometimes it follows the lower levels in patterns, sometimes it essentially forms its own little block, oblivious to the trends on lower levels.
  • Trading cards only seem to have a mild effect. Polywar and Blaite have similar backgrounds in most aspects, yet their numbers are similar. Same goes for Finnish Roller and Control Craft 3. There is some bump in entry numbers, but not too significant.
  • Artifex Mundi HOGs are pretty popular in the user base. Those that were available the cheapest and in the most convenient ways have clearly lower numbers even on level 0, including the only game that managed to flatline after level 7 with zero entries over almost a working week.

I also did another experiment, this time by posting 120+90 keys, mostly on level 0 and level 1, where we suspect most of the auto-entry script users and bots converge. This experiment used delayed starts on giveaways: most went live on the site 12-36 hours after creation. I cross-referenced the numbers to another set of fully open ones where the timing was more traditional, meaning it started within a few minutes to an hour after creation.
The result is that I can safely say that script users and bots mostly prowl the new giveaways page. On the "instant" ones, it took only a few hours to get 2-500 entries, with another 600–1000 entries arriving within the last three hours. The entry number difference between instant and delayed was sometimes close to double in the instant's favour.
So, if you wanted to avoid at least a good portion of the bots, you just had to delay your giveaway by at least a day or more. But this is going to be apparently fixed with the next site update, according to the admin, cg.

Additionally, this large amount of public ones confirmed what I always thought, that those who use Easy SteamGifts or other one-click entry methods mostly enter things that are on the first few pages, meaning the ones that are about to end. Some of these methods offer a pre-defined message along with the entry, and I have received up to a hundred messages from a handful of people in the last two hours before the giveaways ended, but never before.

7 years ago*

Comment has been collapsed.

Thanks for the data. You could also add an info if the games have cards or not.

7 years ago
Permalink

Comment has been collapsed.

I was thinking the same. I seem to recall a script that would let you know that a giveaway has card drops and I think that's a major influence on what some people enter.

7 years ago
Permalink

Comment has been collapsed.

Yes, SG game tags can do that.

7 years ago
Permalink

Comment has been collapsed.

Ah yes. Thanks for the link but I think I'll pass on installing that one so I don't get inspired to enter games just for the cards (weak willpower).

7 years ago
Permalink

Comment has been collapsed.

It would make the first column unreadable. Even the bundled status was pushing it.

7 years ago
Permalink

Comment has been collapsed.

Yes, but you could do a trick... for example, to write ratings for games with cards bold or italic. :)

7 years ago
Permalink

Comment has been collapsed.

View attached image.
7 years ago
Permalink

Comment has been collapsed.

Partly. It also spanned two months by now and it is still going, so it is rather spread out in time. :)

7 years ago
Permalink

Comment has been collapsed.

I see you still rocking them gifs

7 years ago
Permalink

Comment has been collapsed.

Thanks for doing all that work and providing the results. Will you be sharing re-roll data too?

7 years ago
Permalink

Comment has been collapsed.

Nah, too much looking through stuff. Rerolls were a lot more prominent in the second experiment, the 120+90 pack.

7 years ago
Permalink

Comment has been collapsed.

Deleted

This comment was deleted 5 years ago.

7 years ago
Permalink

Comment has been collapsed.

yeah, I sit on the new page too, and have noticed that some GAs still don't show there, despite this previously noted update:

April 15, 2016

  • New is now sorted by start time, instead of created time.
7 years ago
Permalink

Comment has been collapsed.

I think that's a bug, it looks like the New page is still only showing giveaways created in the past 24 hours. This means if a giveaway was created 2 days ago, but started right now, it would not appear at the top of New. I'll fix this in the next update that will likely go out today or tomorrow, and have the New page show all open giveaways, sorted by their start time.

7 years ago
Permalink

Comment has been collapsed.

ou now you take away a cool anti-bot trick ^^

can the real start time be shown on GAs too?

7 years ago
Permalink

Comment has been collapsed.

It could, but giveaways are already crowded with information, so we need to decide what information is most important, or find ways to better organize the information that is already being displayed.

7 years ago
Permalink

Comment has been collapsed.

Currently creation time is always shown, but for me seems the least important.

Why not always have start (on the right) and finish (on the left) times displayed?

Creation time can be in a tooltip of start time.

Edit, some more details:

For example changing tooltip from "February 23, 2017, 2:38pm" into "Created: February 23, 2017, 2:38pm; Started: February 23, 2017, 3:00pm"

Currently, once a GA runs, start time is gone for the user. Before it started, finish time is invisible. I always considered that a lack of useful info

oh and I'm honored to have gotten a cg reply :3

7 years ago*
Permalink

Comment has been collapsed.

Thank you for posting your experiments data Talgaby. :-)

I guessed correctly, in one of your giveaways, that in the 120 keys part you would have 20 re-rolls but can I ask how many re-rolls you had in the 90 keys part?

7 years ago
Permalink

Comment has been collapsed.

That one is on day 2 of 3. 4 requests, 3 approved, 1 permaban so far.

7 years ago
Permalink

Comment has been collapsed.

That sounds much better than the 120 part results (19 re-rolls approved out of 20 requests and a couple of permabans).

7 years ago
Permalink

Comment has been collapsed.

Now give us those charts!

7 years ago
Permalink

Comment has been collapsed.

Uhm. It is right there…

7 years ago
Permalink

Comment has been collapsed.

I mean, like Excel ones. Pies etc.

7 years ago
Permalink

Comment has been collapsed.

I cannot find a free interactive online one I can actually get a quick grasp on how to use. I tried a line graph in Excel, but you cannot see anything in it, there are too many lines.

7 years ago
Permalink

Comment has been collapsed.

I made some figures quickly and the results basically follow what talgaby said. There are some assumptions in the plotted figures, which might not be correct though.
1) For x-axis, I assume that whenever each time the game is bundled, the game rating will drop 90%. For example,
POLYWAR is bundled 0 time, so the rating remain 79%.
Grim Legends 3: The Dark City is bundled 2 times, so the rating is 92%0.90.9=74.52%.
2) For y-axis, I normalized the entries to the maximum entries in each level. For example,
I normalize LV0 to 2305 and LV10 to 73.
The following are the results:
LV0 VS LV10:
LV7 VS LV10:
LV5 VS LV10:
It is clear to see LV10 enter rate drops quickly after the modified rating less than 70%. LV7 basically follow the same rule as LV10 and shows slightly higher selectivity than LV5.
I know this won't be accurate and I am happy to re-plot the figures if anyone has any good suggestion.

View attached image.
View attached image.
View attached image.
7 years ago
Permalink

Comment has been collapsed.

At least you tried :D

7 years ago
Permalink

Comment has been collapsed.

this looks really interesting

7 years ago
Permalink

Comment has been collapsed.

Thank you for sharing

7 years ago
Permalink

Comment has been collapsed.

That must be why I keep seeing giveaways popup on the giveaway list that I missed. I have over 6000 games blocked and have more points than I need, so I go through the entire giveaway list (usually 3 or 4 pages) and enter everything I want. Then I only use the new giveaways page to check for new giveaways, but when I go back to the main giveaway page there are giveaways there that did not show up on the new giveaways page, it must be because people are delaying the start time and they are not showing up on the "new" list.

Edit: I'm always worried about entering giveaways too soon because I use the "new giveaways" list. I have on many occasions entered a giveaway withing the first 5 or 10 seconds of it being created and then when I saw that it was just created, I remove my entry and wait a minute or 2 before entering because I'm worried they will think I might be using a bot if I enter too quick. I'm pretty sure bots don't enter immediately like that, but the people creating the giveaways might think they do.

7 years ago*
Permalink

Comment has been collapsed.

Yep, and it could have been deliberate. For me, it sure was. I am a but sad that cg will fix this loophole, since I think too many people use it for point storage only. (Even before the bots apparently called dibs on it.)

7 years ago
Permalink

Comment has been collapsed.

Very interesting experiment talgaby. It's really fun to see how people here on the forums keep coming up with new stuff to try.

7 years ago
Permalink

Comment has been collapsed.

This is a great topic! Bump again!

7 years ago
Permalink

Comment has been collapsed.

Fascinating results. Thanks for taking the time to post them. One important takeway is that game rating is almost irrelevant on public giveaways. I had suspicion that this is the case, and it's good to see it validated with data. I guess I'll keep on giving crap games in my advocacy public GAs :-)

I'm also doing some experimentation (on a tiny scale) in this thread to see how common (if at all) are bots / scripts entering giveaways posted in the forum. So far it seems like a complete non-issue, as you can see for example in the two giveaways posted here.

7 years ago
Permalink

Comment has been collapsed.

Not yet anyway. The few threads known for hosting forum giveaways like the you are not alone one has a few scripts monitoring them. This is why I exclusively post SGTools ones there, those are only leaked rarely (mine seem to be especially rare now, only individual stragglers).

7 years ago
Permalink

Comment has been collapsed.

Was it proved beyond doubt that these threads are monitored by scripts? If so, then I shouldn't rush recommending people to skip obfuscation...

7 years ago
Permalink

Comment has been collapsed.

Not beyond doubt, no. It would be difficult to prove it. Not to mention that with these, the problem is usually the ones who post the links on other sites or forums. Right now the only way to catch those is to use SGTools or some method where you can get a list of those who entered through some other 3rd party means that list who got the link.

7 years ago
Permalink

Comment has been collapsed.

Right now the only way to catch those is to use SGTools or some method where you can get a list of those who entered through some other 3rd party means that list who got the link.

And even that method is flawed unless the people leaking the links are pretty stupid and are leaking them the wrong way.

7 years ago
Permalink

Comment has been collapsed.

SGTools only gives links after the check, so it is difficult to fake it. What baffles me is when people would actually pass the check but use a leaked link anyway.

7 years ago
Permalink

Comment has been collapsed.

From what I've seen so far many people involved in leaking links (or using leaked links) have a very positive ratio (both sent : won and Real CV; granted 80%+ is from region restricted GAs but still) so it doesn't seem to be as much about the profits as about the time.

7 years ago
Permalink

Comment has been collapsed.

You're saying there's a way to bypass the gate and still be considered a valid entry?

Sharing the SGTools link is not leaking, unless it was hidden behind a puzzle.

7 years ago
Permalink

Comment has been collapsed.

The best way to win the game is to not play it.

Even if sharing the SGTools gateway is technically not considered leaking it achieves the same purpose.

7 years ago
Permalink

Comment has been collapsed.

Not if the gate has rules that prevent anyone from joining, which is very commonly the case.

7 years ago
Permalink

Comment has been collapsed.

The weird thing (but maybe that's just my personal experience) is that many people I've seen involved with leaked GAs have a strong ratio (due to deals from the russian Steam Store) that would allow them to pass most SGTool fiters anyway... So I always assumed it's more about the time for those people.

7 years ago
Permalink

Comment has been collapsed.

Early test results hint that bots haven't picked up on my thread yet. I guess I need to give better games ;-)

7 years ago
Permalink

Comment has been collapsed.

Just one question. Was the time-frame that the giveaways ran for also the same? If they ran during different time-frames that would most probably affect the number of entries.

7 years ago
Permalink

Comment has been collapsed.

They ran around 4 days each, like most of my giveaways. The difference should have been within 12 hours I think. The current 4-game batch has the first large divergence as the start time is spread in a 72-hour window, but they all end the same day.

7 years ago
Permalink

Comment has been collapsed.

"Trading cards only seem to have a mild effect."

It is possible that this is because cards can be added later to any game. For example I make sure to give my gf's acc all extra games I can get my hands on, regardless of whether they have cards or not. Because they might one day get cards.

7 years ago
Permalink

Comment has been collapsed.

Thank you for your experiments, I always like to read the results :-)

7 years ago
Permalink

Comment has been collapsed.

It's always nice to see these kind of things. Thanks for all the effort you have put in.

7 years ago
Permalink

Comment has been collapsed.

Eh, it took more effort to compile the datasheet and to find an online tool to convert it to markdown. Creating x amount of giveaways for the same game is actually pretty easy and fast.

7 years ago
Permalink

Comment has been collapsed.

Filling datasheets is the "hardest" part (you also just reminded me that I still have to post results about my last event happened in november) :v

About markdown conversion, can't you put it in the sheet directly?
I mean, instead of putting data in adjacent cells, you could put them as I did in figure below; drag stuff around and you'll fill it quickly; select needed area, copy and paste :)
That's what I do to quickly update my stats thread.

View attached image.
7 years ago
Permalink

Comment has been collapsed.

I just found a decent online tool that managed to convert the spreadsheet quite nice. I think that should be enough, I don't plan to use too many tables on this site. ^^'

7 years ago
Permalink

Comment has been collapsed.

Can you share a link to that tool?

7 years ago
Permalink

Comment has been collapsed.

Sure, it is one of the early hits on Google: http://www.tablesgenerator.com/markdown_tables

7 years ago
Permalink

Comment has been collapsed.

For your statistics, Im lvl 6 and i entered in all your gas for demon hunter 2, grim legends 3 and enigmatis 3 only because i like these games (not for the cards).

7 years ago
Permalink

Comment has been collapsed.

Cool data, thanks for the freelance research tal :D

7 years ago
Permalink

Comment has been collapsed.

The entries of a group giveaway is nearly three times more than a non-group giveaway. It looks like more people are using autojoin for group giveaways.

View attached image.
7 years ago
Permalink

Comment has been collapsed.

Depends on the groups. It is not so far-fetched to assume that those who use automated methods to enter giveaways have also added themselves to as many public giveaway-oriented groups as possible. Even some closed groups could get some of them. And since you are not the first person to show similar stats, I theorise that either the people or the scripts themselves prioritise group/whitelist giveaways over public ones on the assumption of those getting fewer entries.

7 years ago
Permalink

Comment has been collapsed.

Interesting, thanks for the analysis!

7 years ago
Permalink

Comment has been collapsed.

Updated with 4 more rows. Sadly, nothing particularly interesting was found out apart from an anomaly: the unbundled, high-rated indie game Trashville generated a surprisingly lower interest on levels 0–2 compared to the others, whereas levels 4+ showed a bit higher activity.

7 years ago
Permalink

Comment has been collapsed.

thanks, this is really interesting!

7 years ago
Permalink

Comment has been collapsed.

Very interesting read. Thank you.:)

7 years ago
Permalink

Comment has been collapsed.

Any plan to experiment with invite-only vs public GAs? My experiments (linked above) are on a much smaller scale than yours, so results may not be very telling.

One thing I'm curious about is to find at which level does the reroll rate of public GAs become better than L0 invite-only GAs? My theory is that forum goers are much more likely to read discussions about rule breaking, and as such are less likely to break rules themselves (compared to the typical public GA entrants at the same level). I'd be surprised if Public L3 GAs get less rerolls than forum L0 GAs. I wouldn't even bet on L4 public to be better.

7 years ago*
Permalink

Comment has been collapsed.

That may be an interesting endeavour to look further into sometime in the future.

7 years ago
Permalink

Comment has been collapsed.

I was also curious if you'd noticed any relationship between entry level requirements and rerolls. It might be sensible to look at both reroll requests and rerolls granted (unless you know a way to tell whether a user has served their penalty for a past infraction). I had only been considering public giveaways, but I'd find any analysis, such as the one Yirg mentions, interesting (even a test of the assumptions that forum users learn and follow the rules).

7 years ago
Permalink

Comment has been collapsed.

What I find interesting is that some people who have worked their way up to level 10 enter garbage giveaways (like bold new world). I guess I thought that the higher up the ladder the less number of +1 collectors would be present but that assumption seems to not correlate with the data. I might be doing this site wrong since I only enter games on my wishlist. Maybe I should rethink my approach and/or membership.

7 years ago
Permalink

Comment has been collapsed.

I guess I thought that the higher up the ladder the less number of +1 collectors would be present

The stark opposite. The higher the level, the bigger the hoarding factor gets. There are barely a handful of us on level 8+ with a library under 1000 games, even less under 500 games.

7 years ago
Permalink

Comment has been collapsed.

I can proudly say that despite having a far too big steam library, I did not enter for any of the trash giveaways ;) (in fact, I think I've hidden most of them)

7 years ago
Permalink

Comment has been collapsed.

Updated with six more games. A removed Pong clone generated as much interest as a once-bundled Artifex Mundi HOG. I am starting to be more and more convinced that starting or ending a giveaway on a weekend has an adverse effect on entry numbers.

7 years ago
Permalink

Comment has been collapsed.

Sign in through Steam to add a comment.