Finalized Instructions for the October Contest
Oct 1st 00:00:01 EST, Syncing your kaggle account with your substack email, What to send in the email, The analysis, and Oct 31st 23:59:59 EST
These are the official finalized instructions for the October Contest.
If there are any questions or concerns, leave them in the comments below, or hit me up on twitter: https://twitter.com/BowTied_Raptor
Table of Contents
Oct 1st 00:00:01 EST
Screenshot your kaggle account
What to send in the email
The Analysis
Oct 31st 23:59:59 EST
Step 1 - Oct 1st 00:00:01 EST
The first step of the contest is to wait till Oct 1st 00:00:01 EST. On that specific date/time, I’ll send out a mass email invitation to the October Kaggle data science contest. If you have paid member status, you will automatically get the invite. The email will be titled: BT_Raptor - October Contest Invitation.
You can use the button below to become a paid member:
The invitation email will contain the following:
My email
A link containing the invitation for the actual contest
If you did not get the email:
Check your junk/spam folder, it might be there
Make sure you have added my email to the safe emails, so that it doesn’t get auto blocked
If you still did not get the email, hit me up on twitter, and tell me what email to send to: https://twitter.com/BowTied_Raptor
I’ll do a simple double check to confirm that you do indeed have the paid member status, and then fire off the invitation to you
I will periodically be changing the email invitation link to prevent fraud.
Step 2 - Syncing your kaggle account with your substack email.
Once you have clicked on the email link, you’ll be prompted to either login with your kaggle account, or to sign up and make a new account. Go ahead, and take care of that, you’ll then stumble onto this page here:
In order to know who’s kaggle account is synced with whose email, and where to send the proceeds to the winner, I’ll want to know your substack email, with your kaggle profile. Here’s all you need to do:
Click on your profile in the top right, and just show me a screenshot that looks like this:
This way, I know your kaggle account. Once you have this screenshot ready, just send me an email with the above screenshot, and your substack email, and where you would like the proceeds.
The invitation email will tell you where to email the above information to. For more details see below.
Step 3 - What to Send in the Email
When you send me the email, I just want 4 simple things:
The substack email you are using: Like I said, I will do a simple cross check to see that you do indeed have the paid user status on substack on Oct 1, and Oct 31.
Where you will be storing your report/markdown presentation: My personal recommendation is to just make a private repository on github, and just store your work, and your presentation file there. Here is a link to my github.
Where you would like the proceeds: I noticed not everyone can take paypal, that is fine. If you can’t handle paypal, you can just tell me your ETH address (crypto).
A screenshot similar to the above, but for your kaggle profile: This way, I can track your substack email, and sync it with your kaggle profile, so that way, I know who is who.
Once the contest deadline is hit on Oct 31st, I’ll send a simple message back to the winners and confirm that they have indeed put the correct paypal, or ETH address, and then voila, the winners get the money.
If there are any issues, hit me up on twitter: https://twitter.com/BowTied_Raptor
Step 4 - The Analysis
The Dataset
You are given a simple training dataset called train.csv, which has 2000 rows for you to use for your training data. The target y-value is called: Total Fatal Injuries
Once you have a working model, simply download the test.csv file, and use this to predict your target y-value
Once you have your predictions ready, download, and take a look at the submissions_template.csv. This file tells you what your submissions should look like
Please make sure you have the ID column, and the Total Fatal Injuries when you are submitting your predictions.
Kaggle will use what you have uploaded, and immediately split the predictions into 2 categories: Public and Private
The public scores will give you immediate feedback based off your predictions.
The private scores will be kept private for everyone until the contest ends, and this is what will be used for the actual finalized rankings, and for handing out the prizes.
If you are not happy with your results, no worries, you are allowed a maximum of 5 submissions per day. This gives everyone plenty of time for the analysis, and to maintain the top spot.
Note: I manually altered the data a bit to make it more realistic. It has NAs, 0s, typo errors, capitalization & small letters. Basically, things you will encounter in the real world, had a real company given you this data, instead of the nice and clean datasets you are used to on kaggle.
Holding Onto Your Top Spot
Once you make your submission, you will notice that you are not the only one competing in this contest. It’s possible that today, when you upload your predictions you take the top spot in the public leaderboards. The name of the game is to have the best score on the leaderboards once the contest ends on Oct 31st. This means that you could show up 1st place on the leaderboards at night when you submit your predictions.
Then wake up in the morning to find out you de-ranked quite a bit. The key to winning these type of competitions is to treat it as a marathon, and not a sprint. Keep looking for little nooks and crannies, other research papers, and whatever else you can stumble across that can help you get a little edge. Several little advantages are what you are looking for in order to win this.
Here’s a few helpful guides:
Step 5 - Oct 31st 23:59:59 EST
Once the datetime is Oct 31st, 23:59:59 EST, the contest is officially finished. When it is officially finished, no one will be allowed to upload any more predictions/submissions. One hour later, kaggle will use the other 50% of your predictions for the private leaderboard, and you’ll notice a few ranking changes.
These finalized private leaderboards are the finalized leaderboards once the contest ends. This is why I emphasized the section earlier about continuously trying to improve your model, and score.
Once the official leaderboards are here, I’ll make an announcement, and start reaching out to the winners, and asking for a confirmation as to where they’d like to receive the money, and voila, contest is over.
Prizes
1st Place: 1000 USD
2nd Place: 100 USD
3rd Place: 50 USD