Domain Auction Prediction Algorithm | DomainInvesting.com

Domain Auction Prediction Algorithm

14

Every once in a while, I think of a tool I think would be helpful to my business. I don’t have any type of computer programming background, so instead of spinning my wheels and throwing money at something I wouldn’t be able to adequately tackle, I figured I’d write a blog post about it and maybe someone will create it.

Every domain name that goes to auction, whether expiring auction at NameJet or private auction at TRAFFIC, has many unique attributes. There is extension, keywords, length, inbound links, Google results, keyword search numbers, pagerank, number of other extensions registered, how the domain name was used previously…etc.

All domain names probably have dozens of unique attributes that can be identified, and each attribute plays an important role in the domain name’s valuation.

I would love to have a tool that uses an algorithm to predict the closing price of a domain auction before the auction begins. With thousands of auctions occurring daily across many platforms, there is plenty of data that can be used to predict the sale price. I think it would be a great add on to a tool like FreshDrop, and it could even be a Firefox or Chrome extension that could be enabled on the auction platforms, allowing me to sort by predicted closing price.

This tool is different than automated appraisal tools like Estibot because it would actually predict closing prices rather than value. In addition to a domain name’s attributes, the tool would also take the auction conditions into account as best as possible. For instance, it would be able to use the auction’s start time, venue, number of backorders, high bid prior to auction, and other factors into account when predicting the closing price. With each auction that the algorithm records, it would be able to self adjust for more accuracy.

Do you think a tool like this would be helpful, and would you pay for something like this? How would you go about getting something like this built?


About The Author: Elliot Silver is an Internet entrepreneur and publisher of DomainInvesting.com. Elliot is also the founder and President of Top Notch Domains, LLC, a company that has sold seven figures worth of domain names in the last five years. Please read the DomainInvesting.com Terms of Use page for additional information about the publisher, website comment policy, disclosures, and conflicts of interest.


Reach out to Elliot: Twitter | | Facebook | Email

Comments (14)

    Donny

    I used to have a tool that worked like what you were talking about. The problem is when it’s right, it’s really almost perfect. But when it’s wrong, it’s completely wrong.

    We always put our domains in categories, where you would have a $50k to $100k range. It’s really hard to say a domain will sell for $52315.50.

    June 18th, 2013 at 11:15 am

      Elliot Silver

      Was it a private / in-house tool?

      I would eyeball each domain name I bid on, so an anomalous result wouldn’t necessarily cost me money.

      I would imagine there are plenty of times when an auction sells for less than predicted for a reason that is unpredictable. For instance, I won a domain name for $1k very recently and flipped it to the underbidder who was out of town when the auction ended. I would have bid up to $2,500 and he would have bid more (considering what he paid me), so if the predictor had shown a sale price of $3k, I would have been even more comfortable bidding.

      June 18th, 2013 at 11:19 am

    RaTHeaD

    i know a little about domaining. if you want i can randomly assign a value to every domain name. and i will be way off almost every time but in total i will be close to perfect. something i learned from carl sagan… but don’t quote me as i probably got it wrong as i am usually drunk.

    June 18th, 2013 at 11:37 am

      Elliot Silver

      If I had a dollar for all of the insightful comments you’ve made, I’d probably be able to buy a cup of coffee at Starbucks… maybe even a venti :)

      June 18th, 2013 at 11:38 am

      RaTHeaD

      see… that’s one more thing i like about you. i’m not even sure whether or not that’s an insult.

      June 18th, 2013 at 1:20 pm

      Elliot Silver

      Me neither!

      June 18th, 2013 at 1:20 pm

    todd

    You can’t predict the end price of an auction because you can’t predict what someone is willing to spend. If someone is buying to flip it their price would be much lower than the guy who wants to develop it and his price would be lower than if an end user is in the auction. Its unpredictable. I don’t think this is possible on keyword domains but its absolutely possible if you narrow it down to 3Letters, 4Letters, short number domains etc…..because you are basing the predictions on actual sales data that is extremely similar from one domain to the next. I don’t see how you can do that with keywords because no two are the same.

    June 18th, 2013 at 11:46 am

    Francois

    A pair of months ago we analyzed all the auctions of Sedo for the last 4 years tahnsk to ClosingAuctions.com data and discovered ~50% auctions finished with only 1 bid.

    So now you can easily predict half auctions will never get a second bid.

    June 19th, 2013 at 6:30 am

    Irfan

    Assume this tool gets developed and all Domainers bought this and are seeing the predicted closing prices during auction then what happens?

    1.Up to the predicted closing price bids will be considered as normal or like reserve met by Domainers.

    2.Real bidding gets started soon after the predicted price is met.

    In my view this tool if developed may just serve as a benchmark for bidding.

    June 19th, 2013 at 6:46 am

      Elliot Silver

      I assume if it’s created, 99% of people will be do cheap to buy / use. Also, the nuggets will probably be in the $500-1,000 range, there there are probably hundreds of those daily – too many for most people. This will let users search using their own criteria. I assume most of the names will already be on the radar anyway, but it’s just an extra tool to cross check.

      June 19th, 2013 at 7:07 am

    Mike

    I am an AI specialist and I can tell you that if such a tool is created, even if only otwo persons use it in the same auction, the prices will go way up.

    Ultimately the tool will no longer predict the prices correctly, because its use is affecting the prices.

    @Donny : perhaps your tool was missing an important parameter. That said, I am not implying that it is always feasible to include the missing parameter and fix the problem.

    June 19th, 2013 at 12:05 pm

      Elliot Silver

      Very interesting – thank you for sharing. That’s a big concern of using a tool that others have access to use as well.

      June 19th, 2013 at 12:18 pm

    George

    I agree with Mike.
    There are too many variables in that formula to get the desired results.
    On the other hand, if someone obtains that tool, I doubt that sells the system. I wouldn’t do it. Anyway, is my own personal point of view.

    June 19th, 2013 at 12:57 pm

    Troy Rushton

    Interesting idea EJS, definitely agree with the comments above, a difficult one to get right with so many human variables & even if you could predict, I believe it would then actually create price ceilings, thus eliminating the fun of the auction :-). Conversely, automating bid limits based on variables is certainly more achievable for EMD’s….

    June 19th, 2013 at 10:32 pm

Leave a Reply

Name *

Mail *

Website