GD Auctions API Bot Behaviour

[Brandworthy]

Some information about one particular GD Auctions API Bot and how their automated bidding works. I try not to watch live auctions. Too easy to get caught up in the ebb and flow (and end up overbidding). But I sat and bid some auctions to better understand the programmed behaviour of the bot. All the bidding activity is designed to minimise the cost and maximise the probability of winning. Here are some of the traits it displayed:

1. Initial bid is a backorder. Contrary to what some may think after Godaddy raised backorder pricing some time back, Backorders are marginally cheaper than $5 closeouts when buyers purchase in bulk. The cheapest closeout is $13.47, but when you buy backorders by the hundreds Godaddy will offer you a price around $0.20 cheaper than this. Possibly cheaper if you buy thousands of backorders (which I suspect teh ownerof this bot likely does).

2. Right after the backorder is created a second bid is placed at the next bid amount (i.e. $5 higher). This behaviour is replicated once the bot becomes the high bidder also (see later). I suspect this second bid is to retain the high bidder position if a second bidder comes in with a new non-proxy minimum bid. So Bidder 1 (the bot) places backorder at $10. Immediately after Bidder 1 also places a bid at $15. If a new Bidder 2 places the minimum bid at $15, Bidder 1 is still the high bidder at $15.

3. Approximately 20 seconds before the auction is set to end the bot checks the current bid. Becasue GD's web interface only shows hours and minutes it's not possible to see if this timing is set or varies. If the bot is still the high bidder I suspect they do not check again (very difficult to check via web). If they are NOT the highest bidder they will place a single bid $5 higher than the current high bid. The API response will tell them if their recent bid is the high bid or not. If not, they will immediately place a new bid $5 higher, and continue doing so until they are the high bidder (or their bid budget is reached, more on this later). In the web interface you can clearly see this behaviour when there are a very high number of bids made in less than a minute, which would be impossible to achieve manually.

4. If the bot's bid becomes the high bidder a secondary bid $5 USD higher is immediately added, similar to (2) above, subject to the bots bid budget. This is done to minimise the price at which bot finally becomes the winning bidder (no need to bid $5 more if a single higher bid is made).

5. At no time will the bot make a proxy bid. This suggests that the bot has been programmed to take as much time as possible. This strategy may be designed to wear out human bidders, and in fairness, a part of me admires the tactics used here. But as a human bidder, I imagine most people will despise this behaviour.

6. The bot does have a bid budget, but I don't have enough data to posit how this might be set. Right now I suspect it might be set at 8-12% of an appraised value.

Based on what I've seen this bot is optimised in multiple ways to acquire domains for the lowest possible price. From using backorders to setting secondary bids at minimum next bid level, and not placing proxy-bids to adding new challenging bids at the latest possible point in time, the bot is designed to minimise costs, maximise duration, and win as many auctions within budget as possible. Placing a secondary bid $5 above the bots current winning bid is very clever IMO, as this negates at least some cases where you would have to bid an additional $5 to finally win the auction.

Hope the above might be useful to some GD auctions users who come up against this opponent!

 

[wwwweb]

Bidding at godaddy is just a waste of time these days to be honest, I got better things to do, than let bots sit there and bid $5 for 2 hours straight.

 

[Michael M]

...So, the bot does not necessary maintain "closed" list of domains he wants, but is programmed to extend their wishlist based on our interests. It is was shown in related threads that, even though expiring domains with 1 bid are not included into "most popular" webpage search results, they are reported to api bots.

P.S. The above behavior may be of another bot (or bots), not the one @Brandworthy wrote about

A bot can easily perform an advanced search and look at all items with bids. This will alert them of domains that humans are interested in that the bot may not identified initially, but it can decide is valuable later by a gauge of auction interest.

 

[Michael M]

Some information about one particular GD Auctions API Bot and how their automated bidding works. I try not to watch live auctions. Too easy to get caught up in the ebb and flow (and end up overbidding). But I sat and bid some auctions to better understand the programmed behaviour of the bot. All the bidding activity is designed to minimise the cost and maximise the probability of winning. Here are some of the traits it displayed:

1. Initial bid is a backorder. Contrary to what some may think after Godaddy raised backorder pricing some time back, Backorders are marginally cheaper than $5 closeouts when buyers purchase in bulk. The cheapest closeout is $13.47, but when you buy backorders by the hundreds Godaddy will offer you a price around $0.20 cheaper than this. Possibly cheaper if you buy thousands of backorders (which I suspect teh ownerof this bot likely does).

2. Right after the backorder is created a second bid is placed at the next bid amount (i.e. $5 higher). This behaviour is replicated once the bot becomes the high bidder also (see later). I suspect this second bid is to retain the high bidder position if a second bidder comes in with a new non-proxy minimum bid. So Bidder 1 (the bot) places backorder at $10. Immediately after Bidder 1 also places a bid at $15. If a new Bidder 2 places the minimum bid at $15, Bidder 1 is still the high bidder at $15.

3. Approximately 20 seconds before the auction is set to end the bot checks the current bid. Becasue GD's web interface only shows hours and minutes it's not possible to see if this timing is set or varies. If the bot is still the high bidder I suspect they do not check again (very difficult to check via web). If they are NOT the highest bidder they will place a single bid $5 higher than the current high bid. The API response will tell them if their recent bid is the high bid or not. If not, they will immediately place a new bid $5 higher, and continue doing so until they are the high bidder (or their bid budget is reached, more on this later). In the web interface you can clearly see this behaviour when there are a very high number of bids made in less than a minute, which would be impossible to achieve manually.

4. If the bot's bid becomes the high bidder a secondary bid $5 USD higher is immediately added, similar to (2) above, subject to the bots bid budget. This is done to minimise the price at which bot finally becomes the winning bidder (no need to bid $5 more if a single higher bid is made).

5. At no time will the bot make a proxy bid. This suggests that the bot has been programmed to take as much time as possible. This strategy may be designed to wear out human bidders, and in fairness, a part of me admires the tactics used here. But as a human bidder, I imagine most people will despise this behaviour.

6. The bot does have a bid budget, but I don't have enough data to posit how this might be set. Right now I suspect it might be set at 8-12% of an appraised value.

Based on what I've seen this bot is optimised in multiple ways to acquire domains for the lowest possible price. From using backorders to setting secondary bids at minimum next bid level, and not placing proxy-bids to adding new challenging bids at the latest possible point in time, the bot is designed to minimise costs, maximise duration, and win as many auctions within budget as possible. Placing a secondary bid $5 above the bots current winning bid is very clever IMO, as this negates at least some cases where you would have to bid an additional $5 to finally win the auction.

Hope the above might be useful to some GD auctions users who come up against this opponent!

I believe there are several entities running bidding bots at GD auctions, but I have seen this one you are speaking about. I believe this is the most active and sophisticated bot running currently at this venue.

 

[tonyk2000]

I've been told that's against their TOS

Unfortunately there is no evidence that the TOS are the same for all

 

[wwwweb]

I was watching two last week $5+ bids ran for over 2 hours past when the auction supposed to end both were won by same company

Level 2 LLC
[email protected]
internet traffic nameservers

Looking further into this it seems this persons namejet username is markmon

There is some allegations I found via online search that have been posted in the comments of this blog below:

https://onlinedomain.com/2018/04/10...-this-time-with-an-oliver-hoger-owned-domain/

 

[MapleDots]

It is getting more difficult to win domains across all platforms that is for sure.

I see only a few solutions....

- Bid only on domains you really want and have a potential end user for.

- Bid your highest amount and hope the aftermarket companies deem that too expensive.

Personally I am being quite picky in what I bid on and I look at REGULAR auctions to see what the owner is asking but I then contact through whois and make an offer direct reminding the lister that there are no fees when dealing direct. I landed two this week after reminding the owner of the domains there was a 20% fee. We were only about 10% apart in our negotiations when I reminded him that at auction he has to pay 20% so if he takes my deal he is 10% ahead. That goes with any private auction, you can right away offer 20% less to make up for the auction fees. Why give that away when it can be retained by the buyer/seller. In the last two weeks I have retained $3,600 that would ordinarily have gone to the auction house.

Of course that cannot be done on expiring auctions but I have pretty well determined unless you bid crazy amounts you cannot win against the bods.

Save your time and energy and be selective in what you are looking for. Determine if the potential domain suits your portfolio and don't be shy to send an offer to the owner. If the owner has privacy activated remember you can still contact them through the privacy email. Be direct, tell them who you are and negotiate from there.

 

[tonyk2000]

Someday somebody may develop an anti-bot, that will:

- detect domains overvalued by Godaddys appraisal system

- place an initial bid on these domains, without any real intention to win, just to be outbid by the bot in question

- add +5 bids in a last minute, two or three times - to help human "watchers" of the most active auctions feed on website, as they will notice and join

- stop at this point, as those human watchers will bid up the bot in question up to its limits (or win the domain themselves in some cases, which should be then considered as a side effect of anti-bot technology)

Yes a fantasy of course, but not impossible... Will be interesting to watch bot wars in this case

 

[UXela]

1. If you know the exact algorithm it is fun to watch this bot win crap domains for its max budget;
2. This is NOT HD's bot (at least not the one I played with);
3. This bot does NOT use the API (at least not directly);

 

[tonyk2000]

This bot does NOT use the API

how did you find this?

 

[Brandworthy]

You can make use of the API through a third party.

I have API access, and I've been told that's against their TOS. If you know of third parties offering this I suggest you reach out to GD and tell them about it.

There is (at least) one way to do what this bot is doing without API access.

I'd like to hear what that is, please? You can indeed post a bid without API access, but in order to appraise if that bid is the highest bid, you would need to request the auction page. There is no way that the page can be requested, served and parsed multiple times in a single second without a multi-threaded set-up, and even then I do not think there'd be any way to submit 10+ bids within a second using any other method than their API.

I have used automated tools which submitted bids via the web interface, but no such tool could possibly submit 10+ bids and verify the response in a single second. I'm honestly interested if you know of some alternative, so please share.

 

[UXela]

I have API access, and I've been told that's against their TOS. If you know of third parties offering this I suggest you reach out to GD and tell them about it.

Sure - the second GD becomes an ethical company and the playing field becomes leveled on GD auctions I'll start sharing info with them free of charge. I'll also teach their incompetent coders how to code and remove the plethora of security glitches, etc.

In the meanwhile I'll stick to the opinion that:

• GD's business practices regarding auctions are unethical => they don't deserve to be helped FoC;
• GD is an auctioneer but does not follow auctioneer laws => they don't deserve to be helped FoC.

Also, in the meanwhile I will continue to use GD auctions the best I can while following the rules even if I strongly believe that the same rules don't apply to all auction participants.

You can competitively bid on GD auctions without API access and do that without breaking ANY of their rules. It just takes a lot of time writing some code.
You don't even need to access results from their API through a third party. That just adds latency and basically wastes your time.

I'd like to hear what that is, please? You can indeed post a bid without API access, but in order to appraise if that bid is the highest bid, you would need to request the auction page. There is no way that the page can be requested, served and parsed multiple times in a single second without a multi-threaded set-up, and even then I do not think there'd be any way to submit 10+ bids within a second using any other method than their API.

I have used automated tools which submitted bids via the web interface, but no such tool could possibly submit 10+ bids and verify the response in a single second. I'm honestly interested if you know of some alternative, so please share.

You don't need to time things to the 0.01 second. You are looking at it the wrong way.
The sync is the key, not the timing. Think of it as priority vs. time.

 

[wwwweb]

I was watching two last week $5+ bids ran for over 2 hours past when the auction supposed to end both were won by same company

Level 2 LLC
[email protected]
internet traffic nameservers

Looking further into this it seems this persons namejet username is markmon

There is some allegations I found via online search that have been posted in the comments of this blog below:

https://onlinedomain.com/2018/04/10...-this-time-with-an-oliver-hoger-owned-domain/

Funny how one thing uncovers another:

https://www.namepros.com/threads/faheem-chapnames-level2-scam.944946/

 

[MapleDots]

But couldn't an auction site ban the seller if he just pulls the domain off during the auction? If he already received some bids, they are legally binding.

Yes, you have to make sure there are no active bids. Most of the ones I do that for will have no bids anyways because they are priced too high. Right away you can negotiate by 20% because of the auction fee. So if the guy wants 5k you remind him he only gets 4k and we can do the deal at that if we go private.

 

[Brandworthy]

You don't need to do that sequentially. Key word: multithread. That's how all those tools work.

You're quite obviously missing the part about needing to parse the result in order to decide what to do next. If you want to just fire over 10 bids that's fine (and dumb), but if you want to send over 10 bids because none of the previous bids made you the highest bidder then you MUST do this sequentially.

And yes, it does matter if you're a BH or WH. If you are WH, you may look at this kind of stuff amazingly, and keep repeating "but this can't work, if you understand how HTTP works, bla bla". But if you are BH, you just use the tools like this daily, and if you need go and order a bot for your needs from a coder.

I posted this hoping it might be useful to others. But if you haven't a clue about how this works or what you're doing it's probably best just to ignore this thread.

 

[wwwweb]

I am also wondering has the increased bot activity anything to do with increased number of expired domains that received their first bid noticeably before (days before) the end time. Domainers always used to prefer to place bids in last 5 minutes, or, at least, not in days.

So are these bids all enduser bids? Unlikely.

The only explanation I ended up with so far is somewhat "irrational":

If somebody has a list of domains he might want to acquire, and, due to increased bots activitly, will unlikely acquire (not paying normal wholesale prices at least), he probably wants to at least add his 2 cents to decrease the portfolio quality of said bots. The bots will change strategies to something more rational (or shutdown) earlier or later anyway. Even hugedomains now started to drop domains again and is more selective with bidding. Placing first bid early will add more attention, increase sale price to even more irrational levels, and will force bots to pay more. Which fact an early human bidder might consider a positive by itself in a long run.

Any other explanations why would a domainer place an early 1st bid (so routinely)? Why would anybody want to "subscribe" _not_ to win the domain?

All good questions, there is a lot of irrational strategies employed in the current aftermarket bidding strategies, which has many scratching their heads. Some people are just willing to make infinite amounts to win an auction, all these automated systems, playing on the frustrations of human elements is a perfect storm, and a great win for Godaddy.

What is the purpose of these people buying these domains, they are not using them, they are buying them to relist them to potential buyers, but if you end up paying to much, and out pricing your buyer what is the point? We all hear what most end users say, I will find a cheaper alternative etc... The logistics of current bidding strategies do not make much sense, These are the names we bought years ago for under $50, most are still sitting in out portfolios. Any downturn in the economy, and things go quiet very quickly, and expensive inventory does not usually give a good rate of return sitting there with an overpriced sticker on it. Anyone who has a decent portfolio, is smart to take money off the table, and let these high biddders sit on them for a few years, and wait for them when they need money. That is why I see guys like D Naidu selling on namejet now.

 

[tonyk2000]

Also: if a human placed the first bid ($12), most frequently within the last 5 minutes or so, the bot is notified of this event within approx. 4 minutes, and it will then join the auction in many if not all cases. So, the bot does not necessary maintain "closed" list of domains he wants, but is programmed to extend their wishlist based on our interests. It is was shown in related threads that, even though expiring domains with 1 bid are not included into "most popular" webpage search results, they are reported to api bots.

P.S. The above behavior may be of another bot (or bots), not the one @Brandworthy wrote about

 

[Brandworthy]

This bot does NOT use the API (at least not directly);

I'm pretty sure the behaviour I observed is not possible to achieve via the web interface. It's simply not possible for any client to make the number of kids I saw in a single second via the web frontend, so I suspect your experience must be with another application making bids. There also no way to "indirectly" use the API.

 

[wwwweb]

The same thing is going on with aicontrol.com right now...over $3k now... with min $50 bids by whatever bot or terminator, I don’t know.

 

[UXela]

You can make use of the API through a third party.
There is (at least) one way to do what this bot is doing without API access.

I don't have experience with other applications making bids. I have coding and networking experience.
Anyway - you don't have to believe me nor take my word for it.

 

[Kate]

Even without API access it's still possible to achieve the same result with some programming.
You can do a lot scripting your own headless browser. The tasks are fairly simple:

• keep track of domains of interest and current bids
• (re)loading pages
• parsing HTML contents
• submitting POST requests (= the bids)
• the thing that needs a bit of work is the timing

Dragging the auctions is a form of attrition. Automation provides an advantage and an edge over your competitors. But you still need a budget.

I wouldn't fixate on the API too much. I'm sure that if you are big spender GD can work out something for you. If they don't, be creative. There is no need to be sitting in front of your desk for repetitive tasks.

 

[Brandworthy]

You don't need to time things to the 0.01 second. You are looking at it the wrong way.
The sync is the key, not the timing. Think of it as priority vs. time.

You said you have Dev and networking experience. You understand that to make a big via their web interface means making a request, and determining if that but it's successful requires review of the response. So please tell me how anything can do that multiple times in a single second. It's physically impossible via the website interface. You might get 1 or 2 bids in a second this way, but not almost 10.

 

[UXela]

@Kate more or less explained it
10 actions / second really not a problem but again - you are looking at it all wrong.
Don't look at it as a blind race. Look at it as completing multiple puzzles while the one closest to completion has the lowest "ring" (in CPU talk).

 

[Brandworthy]

10 actions / second really not a problem but again - you are looking at it all wrong.
Don't look at it as a blind race.

Anyone can blindly fire 10 requests in a second. That's not the problem. The problem is that any subsequent request after the first should only fire if the outcome of the previous bid results in a non-winning position. You cant just fire 10 requests asynchronously. You have to wait for each response to determine whether to bid again. That's physically not possible, or at least exceptionally non-trivial without the API, to achieve 10 times within 1-2 seconds.

 

[UXela]

Let's agree to disagree