Master P and Automated Graded Systems Inc.

[davie]

@wooter

@zorloth

Related to this, the textures of holos from certain sets (i.e., Expedition, Aquapolis) are often naturally “wavy.” That sort of wavy texture, though, could be actual damage on holos from other sets. If the AI is precise enough to detect any sort of surface abnormalities, it will have to take into account what set a card is from. Condition can’t just be evaluated in a vacuum.

@wooter , I’d also be curious to know if the AI differentiates between holo scratches and factory print lines. Basically, can the AI differentiate between factory flaws and handling flaws? And, if so, do they each affect the grade to the same extent?

Yes with each new set we have to update the AI.

The annotations do detect scratches and print lines differently because of the heights. The grade would be affected depending on size and depth of the flaw.

For the time being, are these scans all manually reviewed then? I imagine there will be cases where imperfections are not detected (or scored) correctly, and will need to be labelled manually. It’s only using larger and larger sets of improved labels that you can improve the model, correct?

Also leads into another interesting question: the team must have bootstrapped the model(s) by manually labelling an initial set of scans, yes? And I assume they would continue to do so for the foreseeable future based on your description of how the “AI” improves (with less frequent intervention as the size and quality of the dataset improves)?

[wooter]

@davie

@traineramy

I’m not entirely sure why any startup thinks they would have better tech or better funding than PSA or CGC?

Maybe not funding, but I can see a group of reasonably smart, competent folks coming up with a solution that beats out PSA/BGS with respect to grading consistency. Whether or not that’s enough to take a meaningful slice of the pie is something else.

One thing is for sure: their pricing (especially as a new entrant with no reputation or indication that their solution *is* better) is disappointing. There are loads of new grading companies that have cropped up in the past few years — IMO, touting “AI” and showing off some imaging hardware is hardly enough to differentiate yourself in that environment.

PSA is worth 1B+ and CGC just got acquired for 500M+

If we had this type of funding, we would have 10x the team size and would have our AI ready for all TCGs + sports cards by now. All we can do right now is be the first to bring this AI grading revolution to Pokemon and then add other TCG + sports cards when ready. Hopefully our speed will give us the first movers advantage.

We did raise $1.5m and my partners and I have extra $ from our other ventures. We are also looking to raise $10m which can get us to 10k+ cards per day.

---

In terms of pricing, $20 is very cheap in our opinion. We have 8 subgrades, other companies charge extra. Maybe our higher tiers can be adjusted, I agree on that.

Plus we will be rewarding all of our early adopters with Collector Coin (our crypto) and have new value adds coming soon (NFTs).

[wooter]

@davie

@wooter

1. I don’t think we are promising lower grades. Cards that deserve 10s and are gem mint, get them. We are promising more consistency in the grades along with full transparency on why each card got the grade it received.

2. The more cards we grade the better our AI becomes over time. We do have some key members on our team and are currently speaking with experienced veterans to further enhance our credibility.

3. We actually show many videos on our Youtube of our process and have shown it live at Collect A Con from A-Z. The secret sauce is the patent pending machine that scans the card and our AI models which are constantly updated.

4. Zack, aka Gem Mint Pokemon let us grade this card very early on in June, before we launched. There is 4 subgrades on front and 4 on the back. That averages out to give you the total score.

5. The pop report is robograding.com/pop.

A couple of follow-up questions:

2. In what way does the AI get better? I assume you’re talking about detection of imperfections as you add more and more scans to your internal repository of scans, and not the grade, which relies on some combination of count and severity of these imperfections? If so, can you comment on the current level of fidelity?

3. The videos on YouTube don’t show much at all… and “patent pending machine” and “AI models” are incredibly reductive. Any detail on this?

2. The more times the Pokemon card gets annotated perfectly and uploaded, then the AI will annotate it correctly the next time.

3. The machine we call it AG-1000 because it can scan 1000 cards per day. We are building our 2nd one now. It is our version of the best way to capture the card in a controlled environment. We get a 3D card after scanning it front and back. The laser is a lab-grade laser.

The AI models are constantly updated, 1-2x per month. Before updating all the uploads are manually reviewed which you can imagine is a ton of work.

Think captcha, but instead of captcha it is highlighting a scratch, or a corner bend, or printline, or crease, or whitening, etc.

[davie]

@wooter

@davie

Maybe not funding, but I can see a group of reasonably smart, competent folks coming up with a solution that beats out PSA/BGS with respect to grading consistency. Whether or not that’s enough to take a meaningful slice of the pie is something else.

One thing is for sure: their pricing (especially as a new entrant with no reputation or indication that their solution *is* better) is disappointing. There are loads of new grading companies that have cropped up in the past few years — IMO, touting “AI” and showing off some imaging hardware is hardly enough to differentiate yourself in that environment.

PSA is worth 1B+ and CGC just got acquired for 500M+

If we had this type of funding, we would have 10x the team size and would have our AI ready for all TCGs + sports cards by now. All we can do right now is be the first to bring this AI grading revolution to Pokemon and then add other TCG + sports cards when ready. Hopefully our speed will give us the first movers advantage.

We did raise $1.5m and my partners and I have extra $ from our other ventures. We are also looking to raise $10m which can get us to 10k+ cards per day.

---

In terms of pricing, $20 is very cheap in our opinion. We have 8 subgrades, other companies charge extra. Maybe our higher tiers can be adjusted, I agree on that.

Plus we will be rewarding all of our early adopters with Collector Coin (our crypto) and have new value adds coming soon (NFTs).

I’m with you on this – I think the age of $10 bulk is over. $20 seems like a fair price, and will become increasingly lucrative as the the offering gains traction. Just the $0.02 of an internet nobody – please try to focus on your core product offering and keep the effort alloted on the crypto/NFT cash grab low. The age of web3 is still early and unless there’s a core product reason for doing so, adding “crypto” to any business’ name or an NFT offering might net you a couple bucks from hype chasers, but add zero value otherwise

[davie]

@wooter

@davie

A couple of follow-up questions:

2. In what way does the AI get better? I assume you’re talking about detection of imperfections as you add more and more scans to your internal repository of scans, and not the grade, which relies on some combination of count and severity of these imperfections? If so, can you comment on the current level of fidelity?

3. The videos on YouTube don’t show much at all… and “patent pending machine” and “AI models” are incredibly reductive. Any detail on this?

2. The more times the Pokemon card gets annotated perfectly and uploaded, then the AI will annotate it correctly the next time.

3. The machine we call it AG-1000 because it can scan 1000 cards per day. We are building our 2nd one now. It is our version of the best way to capture the card in a controlled environment. We get a 3D card after scanning it front and back. The laser is a lab-grade laser.

The AI models are constantly updated, 1-2x per month. Before updating all the uploads are manually reviewed which you can imagine is a ton of work.

Think captcha, but instead of captcha it is highlighting a scratch, or a corner bend, or printline, or crease, or whitening, etc.

Got it!

Out of curiosity, has the team built an in-house solution to do this, or do you use a third party vendor (like Scale or MTurk)?

[wooter]

@davie

@wooter

Yes with each new set we have to update the AI.

The annotations do detect scratches and print lines differently because of the heights. The grade would be affected depending on size and depth of the flaw.

For the time being, are these scans all manually reviewed then? I imagine there will be cases where imperfections are not detected (or scored) correctly, and will need to be labelled manually. It’s only using larger and larger sets of improved labels that you can improve the model, correct?

Also leads into another interesting question: the team must have bootstrapped the model(s) by manually labelling an initial set of scans, yes? And I assume they would continue to do so for the foreseeable future based on your description of how the “AI” improves (with less frequent intervention as the size and quality of the dataset improves)?

I have actually advised my team last week to create a blog to share all updates which I think is fundamental. My vision for AGS is to open source and share everything except the machine and the internal code for our AI (but allow API access). Very soon we will be implementing Collector Coin rewards to our users, contributors, annotators and having an open UI/UX to allow anyone to help us with our database, annotations, etc.

As mentioned above, the annotations are reviewed and can be adjusted if something is missed or selected incorrectly. This is really the most crucial step as after every edit, the AI will do a better job after the next update.

Yes we are annotating 24/7. One card can be annotated 10+ times by different people.

[zorloth]

@wooter

@zorloth

Related to this, the textures of holos from certain sets (i.e., Expedition, Aquapolis) are often naturally “wavy.” That sort of wavy texture, though, could be actual damage on holos from other sets. If the AI is precise enough to detect any sort of surface abnormalities, it will have to take into account what set a card is from. Condition can’t just be evaluated in a vacuum.

@wooter , I’d also be curious to know if the AI differentiates between holo scratches and factory print lines. Basically, can the AI differentiate between factory flaws and handling flaws? And, if so, do they each affect the grade to the same extent?

Yes with each new set we have to update the AI.

I’m talking about older sets – for example, does the AI assess cards from, say, Neo Genesis differently than it does cards from Jungle?

The annotations do detect scratches and print lines differently because of the heights. The grade would be affected depending on size and depth of the flaw.

I should’ve clarified – by print lines I meant the print lines that commonly appear on WotC holos, which are physical scratches. Can the AI detect whether or not a holo scratch is a factory or post-factory flaw? And if so, how?

[wooter]

@davie,

we have built upon CVAT and created our own in house solution to share/edit annotations. we are considering 3rd parties but we must adapt our solution to be integrated with theirs.

i can assure you there is no quick money-grabs, everything is with 10 years in mind.

harnessing the power of blockchain/web3 will allow us to differentiate from PSA. it will allow us to digitize these graded collectibles (NFTs), create a free to trade marketplace, and allow collateral based lending/loans. still a bit early to address this as a lot can change but it’s very promising and 100% the future.

[wooter]

@zorloth

@wooter

Yes with each new set we have to update the AI.

I’m talking about older sets – for example, does the AI assess cards from, say, Neo Genesis differently than it does cards from Jungle?

The annotations do detect scratches and print lines differently because of the heights. The grade would be affected depending on size and depth of the flaw.

I should’ve clarified – by print lines I meant the print lines that commonly appear on WotC holos, which are physical scratches. Can the AI detect whether or not a holo scratch is a factory or post-factory flaw? And if so, how?

Right now Neo Genesis and Jungle is graded the same. I know Neo genesis has awful printlines and has more defects but we do not give it any leeway. in the future it might be adjusted, but that will be left up to open-vote by the public.

My bad, I should’ve been more clear. The print lines are differentiated from scratches because the scratch is deeper and is shown as a lower point on the height map in the CSV file. Hope that makes sense.

[davie]

@wooter

@davie

For the time being, are these scans all manually reviewed then? I imagine there will be cases where imperfections are not detected (or scored) correctly, and will need to be labelled manually. It’s only using larger and larger sets of improved labels that you can improve the model, correct?

Also leads into another interesting question: the team must have bootstrapped the model(s) by manually labelling an initial set of scans, yes? And I assume they would continue to do so for the foreseeable future based on your description of how the “AI” improves (with less frequent intervention as the size and quality of the dataset improves)?

I have actually advised my team last week to create a blog to share all updates which I think is fundamental. My vision for AGS is to open source and share everything except the machine and the internal code for our AI (but allow API access). Very soon we will be implementing Collector Coin rewards to our users, contributors, annotators and having an open UI/UX to allow anyone to help us with our database, annotations, etc.

As mentioned above, the annotations are reviewed and can be adjusted if something is missed or selected incorrectly. This is really the most crucial step as after every edit, the AI will do a better job after the next update.

Yes we are annotating 24/7. One card can be annotated 10+ times by different people.

Great responses – I was definitely being a little facetious in some of my comments.

I dig the vision. Sharing some of these views in the public sphere would be great – too many cash grab grading companies and buzzwords to cut through, but it sounds like you have a solid vision, team + plan.

If the team does release a public blog, I look forward to reading it. I think something similar in format to Stitch Fix’s blog, that isn’t too verbose (such that a lay person can get the gist of how things work), can go a long way