Predictive Odds are Questionable

[houghton_hunter]

There are lies, damn lies, and statistics.

Just drew an Elk tag in CO with 10 points, that GOHUNTs predictive odds said was a 0% chance, as they say "based on our simulations we anticipate a 95% confidence level in our predictive odds". As a statistician myself I could have told you that this was silly, its based off of last years data, and really has no way of telling you about application behavior in response to policy changes, season date influence, people who believe in moon phases, as just some of the variables, despite running "simulations". Basically there is no way to know for sure, and without disclosing their model, I probably could have told them there predictions are going to be dead wrong if they only based it on "trends" and don't consider why those trends occur.

Anyone else beat the odds, if you will?

 

[Jbehredt]

Beat the odds by not playing around with gohunt. 60% of the time I win every time.

 

[Jakeb]

Were people actually applying based off of the predictive odds? I took it as an educated guess that’s really no better then my own.

 

[S.Clancy]

95% CI is only based on how many times they run the simulation. They must have been pretty confident with their input variables......

Predictive odds are a joke IMO. Basically, just a marketing ploy.

Interested from folks that have it, did GoHunt predictive odds give the whole range of the CI values or just the one mean value?

 

[houghton_hunter]

Were people actually applying based off of the predictive odds? I took it as an educated guess that’s really no better then my own.

I am sure someone, or probably someones did.

 

[cuerro viejo]

95% CI is only based on how many times they run the simulation. They must have been pretty confident with their input variables......

Predictive odds are a joke IMO. Basically, just a marketing ploy.

Interested from folks that have it, did GoHunt predictive odds give the whole range of the CI values or just the one mean value?

They are just based off past years applications, Ive drawn a 20pt Deer tag with 12pts, in hindsight I should of turned it back, as I was using it as a PP unit anyway

 

[TXtoWest05]

I considered them this year as a data point but ended up on the wrong side of the bell curve both times, CO and Wyoming elk. I was not surprised though. Any person who has looked at Wyoming knows the second year after a price increase is going to cause a massive shift in special vs regular odds as everyone attempts to beet the odds. The key is that 95% of the other players are also seeing the same opportunity. Example unit in Wyoming with no random special tags until a tag increase last year caused a random tag to appear. Go hunt predicted a 5% odd but in reality it had even lower odds as everyone flocks to the unit hoping to be the lucky one.

It seams like the main thing they were doing was just showing people who could not comprehend point creep, point creep. I appreciate them for centralizing the start of my research but there will always be to many variables. They don’t even re run last year’s data for known variables changes like tag increase or reduction which would be the simplest start to a predictive odds analysis.


Sent from my iPhone using Tapatalk

 

[displacedtexan]

My dad's brother had a PhD in statistics, and taught them.

He liked to say that his favorite thing about statistics was that you could make them say whatever you wanted.

 

[ThreeOhEight]

I like GoHunt to an extent that it’s easy to bring up last years data and see summaries of the amount of public land, last years harvest success, terrain, species within the unit, etc.

Really helps me with out of state applications.

 

[LongWayAround]

@houghton_hunter what was the minimum number of points to draw forecasted at?

Are you a resident or not?

 

[houghton_hunter]

@houghton_hunter what was the minimum number of points to draw forecasted at?

Are you a resident or not?

It was forecasted as "guaranteed" at 12 pp, as a resident.

 

[dave_gohunt]

Hey everyone,

Just wanted to jump in real quick, address a few things, and be available for conversation.

Totally understand the skepticism around predictive odds—no system is perfect. That said, Colorado is in a unique spot right now. With major changes coming in 2028, there’s already been a noticeable shift in application patterns. Some hunts are seeing increased pressure, and it's likely that trend continues, though how much is still anyone’s guess. We’ve always been upfront that predictive models aren't a crystal ball. They’re simply a tool to help add context and guide decisions—not a replacement for personal judgment or real-time info.

We’re a feedback-driven company and are constantly improving. This was our first rollout of predictive odds, and while we've hit our confidence intervals in most other states, Colorado is just a tougher nut to crack right now. Still, we’re confident that with continued work, this part of the platform will only get better.

Thanks for all the input!

 

[wapitibob]

Ron Wolds predictive odds for OR are pretty good but he has been doing it for 20 years.

 

[AZ_Hunter_2000]

Once I drew AZ Unit 10 early rifle bull elk with 2 points; at the time it took 19 points to get drawn.

 

[Maverick1]

There are lies, damn lies, and statistics.

Just drew an Elk tag in CO with 10 points, that GOHUNTs predictive odds said was a 0% chance, as they say "based on our simulations we anticipate a 95% confidence level in our predictive odds". As a statistician myself I could have told you that this was silly, its based off of last years data, and really has no way of telling you about application behavior in response to policy changes, season date influence, people who believe in moon phases, as just some of the variables, despite running "simulations". Basically there is no way to know for sure, and without disclosing their model, I probably could have told them there predictions are going to be dead wrong if they only based it on "trends" and don't consider why those trends occur.

Anyone else beat the odds, if you will?

Using predictive odds? That’s about the same as using your rear view mirror to drive your car forward. Going to eventually end up in the ditch, complaining like a little b!tch, because you didn’t draw a tag…..

Is it that time of the year where the “land of make believe” threads get started about how to solve all the tag drawing systems and their associated “problems”?

 

[180ls1]

Across my group they were more helpful/accurate than not.

 

[Jstumbaugh]

We used the predictive odds to help us make our application bet. We had our safety net application and gambled on where point creep would go in on another. We were 4/4 on the hedge and 2/4 on the gamble.

 

[LongWayAround]

Using predictive odds? That’s about the same as using your rear view mirror to drive your car forward. Going to eventually end up in the ditch, complaining like a little b!tch, because you didn’t draw a tag…..

Is it that time of the year where the “land of make believe” threads get started about how to solve all the tag drawing systems and their associated “problems”?

He did draw the tag...

Which makes this post even more curious. Drawing with 10, when 12 should have been the lock isn't even much of a stretch.

 

[LongWayAround]

Once I drew AZ Unit 10 early rifle bull elk with 2 points; at the time it took 19 points to get drawn.

I would assume that you drew in the random portion of the lottery cycle.

 

[5MilesBack]

Is that tag part of the hybrid draw process? Lots of folks draw with that hybrid draw with less points than needed.